Silver Bullet or Ricochet? CEOs’ Use of Metaphorical ......Silver Bullet or Ricochet? CEOs’ Use of Metaphorical Communication and Infomediaries’ Evaluations Andreas König, Jan

Silver Bullet or Ricochet? CEOs’ Use of Metaphorical Communication and Infomediaries’ Evaluations

Andreas König, Jan Mammen, Johannes Luger, Angela Fehn, and Albrecht Enders

Journal article (Publisher’s version)

Please cite this article as: König, A., Mammen, J., Luger, J., Fehn, A., & Enders, A. (2018). Silver Bullet or Ricochet? CEOs’ Use of

Metaphorical Communication and Infomediaries’ Evaluations. Academy of Management Journal, 61(4), 1196-1230. https://doi.org/10.5465/amj.2016.0626

DOI: 10.5465/amj.2016.0626

Uploaded to CBS Research Portal in accordance with the Open Access Policy of Academy of Management:

https://aom.org/uploadedFiles/Publications/OA_New-VS-Old_Policy.pdf

Uploaded to CBS Research Portal: February 2020

https://doi.org/10.5465/amj.2016.0626

https://aom.org/uploadedFiles/Publications/OA_New-VS-Old_Policy.pdf

https://research.cbs.dk/en/publications/silver-bullet-or-ricochet-ceos-use-of-metaphorical-communication-

r Academy of Management Journal2018, Vol. 61, No. 4, 1196–1230.https://doi.org/10.5465/amj.2016.0626

SILVER BULLET OR RICOCHET? CEOS’ USE OFMETAPHORICAL COMMUNICATION AND

INFOMEDIARIES’ EVALUATIONS

ANDREAS KONIGUniversity of Passau

JAN MAMMENUniversity of Erlangen-Nuremberg

JOHANNES LUGERCopenhagen Business School and University of St. Gallen

ANGELA FEHNUniversity of Bamberg

ALBRECHT ENDERSIMD International

We combine literature on rhetoric and socially situated sensemaking to illuminate thechallenges that emerge when chief executive officers (CEOs) try to influence infome-diaries by using metaphorical communication—figurative linguistic expressions thatconvey thoughts and feelings by describing one domain, A, through another domain, B.Specifically, we theorize that because different infomediaries are situated in differentthought worlds, CEOs’ use of metaphorical communication has contradictory effects onjournalists’ and securities analysts’ evaluations: while it triggers more favorable state-ments from journalists, it prompts more unfavorable assessments from analysts.Moreover, we integrate findings from cognitive psychology to argue that these contra-dictory effects increase the more a firm’s performance falls behind market expectations.Our hypotheses find support in an extensive analysis of 937 quarterly earnings calls inthe U.S. pharmaceutical, hardware, and software industries, and of journalists’ state-ments and analysts’ earnings forecasts and recommendations. Our novel theorizing andfindings suggest that the use of discursive frames, especially in the form of metaphoricalcommunication, in firms’ interactions with critical audiences creates thought-provokingand thus-far neglected dilemmas. In developing and testing these thoughts, we con-tribute to and link ongoing conversations in management science, especially discussionsof organizational reputation, executive communication, and impression management.

We gratefully acknowledge helpful guidance andthoughtful comments from Associate Editor Scott Graffinand three anonymous reviewers. We are thankful to MaryBenner, Jonathan Bundy, Craig Crossland, Angelo Fanelli,Thorsten Grohsjean, Donald Hambrick, Carolin Haussler,Nathan Hiller, Harald Hungenberg, Nadine Kammerlander,Vilmos Misangyi, Timothy Quigley, and Martin Weiss, aswell as to participants at the 2012 Meeting of the StrategicManagement Society in Prague, the 2012 VHB Meeting inBolzano, and the 2013 Annual Meeting of the Academy ofManagement for comments on earlier versions of this man-uscript. Furthermore, we greatly thank Yevgen Aba,Andreas Blank, Falk Bosecke, Katharina Boxleitner,Magdalena Brettner, Beatrix Burg, Christian Faller, TheresaFehn, Alexander Ixmeier, Moritz Jenisch, Imke Krause,

Junheng Liu, Christian Mende, Michael Mikesky, DanielaNitschke, Daniel Nowak, Rainer Oehlmann, VerenaKomander, Heribert de Oliveira Kuhn, Philipp Schmidt,Andre Steinbacher, JoannaWalton, AlexanderWessels, andMichael Wiedermann for research assistance during thisproject. Tina Pedersen, Beverley Lennox, Susan Stehli, andLindsay McTeague provided editing support and LorenzGraf-Vlachy developed text analysis software. We aregrateful for the generous support for parts of this researchfrom the Theo and Friedl Scholler Foundation, Germany,and the Swiss National Science Foundation (grant no. SNF157026).Anearlier versionof this paper received theGlueckBest Paper Award from the Business Policy and StrategyDivision of theAcademyofManagement andwas publishedin the Academy of Management Best Paper Proceedings.

1196

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download, or email articles for individual use only.

https://doi.org/10.5465/amj.2016.0626

One of the vital tasks of a chief executive officer(CEO) is to communicatewith infomediaries, such asjournalists, analysts, and social-movement groups(Fanelli, Misangyi, & Tosi, 2009; Gao, Yu, &Cannella, 2015; Washburn & Bromiley, 2014;Westphal & Deephouse, 2011). Infomediaries informspecific audiences and the broader society aboutfirms and, thereby, strongly influence firms’ socialapproval, reputation, and legitimacy (Pfarrer,Pollock, & Rindova, 2010; Rindova & Fombrun,1999; Shoemaker & Reese, 1996). The CEO’s job inthe firm–infomediary relationship is that of a criticalsensegiver who ensures that the narrative receivedand forwarded by infomediaries is one that is be-nevolent toward the firm and consistent with thegoals of the enterprise as a whole (Fanelli et al.,2009). This job is challenging because details aboutthe firm are abundant, highly complex, and ambig-uous, and a CEO typically has relatively little time tocommunicate directly with infomediaries and guidetheir interpretations (Giorgi & Weber, 2015; Healy &Palepu, 1995). As such, CEOs are hard-pressed notonly to carefully select the details they wish to sharewith infomediaries but also to package that infor-mation in away that reduces complexity, and swiftlyand subtly steers infomediaries’ attention toward apositive interpretation.

Of all of the persuasive techniques included inthe canon of rhetoric (Corbett & Connors, 1998), onedevice appears to be particularly well suited foraddressing this challenge: metaphorical communi-cation. “Metaphorical communication [denotes all]figurative linguistic expressions that convey thoughtsand feelings by describing one domain, A, throughanother domain, B” (Konig, Fehn, Puck, & Graf-Vlachy, 2017: 271), where domain A is typically anunfamiliar, difficult-to-grasp domain of knowledgeand domain B is a familiar, concrete domain ofknowledge (Lakoff & Johnson, 1980). For example,Warren Buffett (1985) explained the closure ofBerkshire Hathaway’s textile operations using meta-phorical communication:

[A] goodmanagerial record . . . is farmore a functionofwhat business boat you get into than it is of how ef-fectively you row . . . Should you find yourself ina chronically-leaking boat, energy devoted to chang-ing vessels is likely to bemore productive than energydevoted to patching leaks.

As highlighted by an extensive body of cognitivelinguistics and organization studies (e.g., Antonakis,Fenley, & Liechti, 2011; Johnson & Lakoff, 2002;Mio, Riggio, Levin, & Reese, 2005; Ortony, 1975;

Thibodeau & Boroditsky, 2011), metaphorical com-munication is a structural form of “discursiveframing” (Cornelissen &Werner, 2014: 183) that canenable communicators to simplify messages, steerreceivers’ awareness, and strengthen receivers’ pos-itive attitudes toward the communicated message(for an overview, see Landau,Meier, & Keefer, 2010).Not surprisingly, therefore, a host of publicationshave advised CEOs to use metaphorical communi-cation to convey messages to important audiences(e.g., Antonakis, Fenley, & Liechti, 2012; Gavetti &Rivkin, 2005; Heath & Heath, 2007), arguing thatit makes “something as complex, impersonal, andabstract as finance or business [. . .] sound simple,human, and concrete” (Leith, 2014).

But is metaphorical communication really so con-ducive to CEOs’ sensegiving toward infomediaries?There is indeed reason to suspect that CEOs and theirfirms might not always benefit from using metaphor-ical communication. Most importantly, cognitivelinguists indicate that metaphorical communicationmight entail critical pitfalls because it comparestwo domains of knowledge (e.g., “business” and“seafaring” in Buffett’s quote above) that are sociallyconstructed and, a priori, never fully correspond(Black, 1962; Ibañez&Hernandez, 2011; Steen, 2011).As a result, metaphorical communication is inher-ently imprecise and ambiguous (for overviews, seeRamsay, 2004, and Shenkar, Luo, & Yeheskel, 2008).This could be crucial in the context of CEOs’ com-munication with infomediaries: although some info-mediaries might prefer less precise but easy-to-graspfiguratively conveyed information, others might re-spondmore favorably to detailed and precise literallytransmitted information. Thus, intricate paradoxescould arise because the same aspects of metaphoricalcommunication that appeal to certain infomediariesmight induce other infomediaries to be more skepti-cal. Given that CEOs’ public communication is re-ceived by multiple audiences (Gao et al., 2015), theseparadoxescouldhavedetrimentaleffectson the firm’sapproval among constituents.

In this paper,we aim to illuminate this dilemmabyasking, How does the degree to which a CEO usesmetaphorical communication affect the favorabilityof different infomediaries’ assessments of the CEO’sfirm? To address this question, we envision info-mediaries as socially situated, “constituent-minded”arbiters (Wiesenfeld, Wurthmann, & Hambrick, 2008),as they make sense of the CEO’s sensegiving throughthe prism of giving sense to their own audiences.Moreover, we expect infomediaries’ assessments tobe biased by their tendency to reduce cognitive effort

2018 1197Konig, Mammen, Luger, Fehn, and Enders

(Taylor, 1981), and by the norms and conventionsof sensemaking and sensegiving in infomediaries’respective fields (Lamin & Zaheer, 2012). From thesepremises, we deduce that different infomediarygroups, because their audiences differ profoundly,develop idiosyncratic preferences regarding CEOs’use of metaphorical communication.

To elaborate on our theory, we focus on two par-ticularly important groups of infomediaries: jour-nalists1 and securities analysts.2 More specifically,we hypothesize that journalists report more favor-ably on a firm the more a CEO uses metaphoricalcommunication because journalists can tap into thefamiliar concepts evoked by such communicationwhen writing for their mainstream audience. In con-trast, we argue that analysts assess a firm less favor-ably the more the respective CEO uses metaphoricalcommunication because analysts are socially situatedin a fact-oriented “thought world” (Lamin & Zaheer,2012: 47), and prefer detailed, unambiguous infor-mation when writing for their investor audience. Byextension, we integrate findings from cognitive psy-chology (Baumeister, Bratslavsky, Finkenauer, &Vohs, 2001) to propose that journalists’ and analysts’biases related to CEOs’ use of metaphorical commu-nication intensify the more the focal firm’s perfor-mance negatively deviates frommarket expectations.

We manually coded 937 quarterly earnings con-ference calls to assess CEOs’ use of metaphoricalcommunication in the U.S. pharmaceutical, hard-ware, and software industries from 2002 to 2011.Wefind support for our hypotheses when testing theeffect of CEOs’ use of metaphorical communicationon 25,415 hand-coded statements from journalists,6,969 analysts’ earnings per share (EPS) forecasts,and 393 analyst buy-hold-sell recommendations.

Our study makes several contributions. First, wego beyond prior work on executives’ strategic pub-lic language (e.g., Fanelli et al., 2009; Guo, 2014;McDonnell&King, 2013;Washburn&Bromiley, 2014)and CEO–infomediary relations (e.g., Westphal, Park,McDonald, & Hayward, 2012) by showing that thedegree to which CEOs use metaphorically structured

discursive frames—rather thanmerely the amount andcontent of information they provide—affects firms’approval among important constituents. Second, weadd to the emerging conversation on the contextual-ized implications of the metaphorical communicationusedbycorporate leaders (Cornelissen,Holt,&Zundel,2011; Konig et al., 2017). In particular, we provide andpreliminarily substantiate explanations for why meta-phorical communication might not be a rhetorical“silver bullet” for CEOs, but can instead ricochetwhenCEOs use it to give sense toward certain types of au-diences. Third, we add a new, rich lens to the recentlyintensifying debate on the potentially paradoxicalrole of rhetoric in firms’ communication with a di-verse set of infomediaries (e.g., Lamin&Zaheer, 2012;Zavyalova, Pfarrer, & Reger, 2016).

CEOS AND INFOMEDIARIES: THE FOCAL ROLEOF DISCURSIVE FRAMING

Research has long highlighted that a central task ofchief executives is to construct meaning for internaland external constituents through verbal communi-cation (Gao et al., 2015; Mintzberg, 1973). Suchsensegiving (Gioia & Chittipeddi, 1991) is particu-larly important in CEOs’ interactions with infome-diaries (Deephouse & Heugens, 2009)—third-partyactors who mediate and broker between firms andtheir external audiences by collecting, interpreting,anddisseminating firm-relatedinformation (Shoemaker& Reese, 1996). Infomediaries are focal addressees ofCEO communication because they strongly influencea firm’s social and economic approval (Zavyalova,Pfarrer, Reger, & Shapiro, 2012), and often exert pres-sure on firms and their leaders (Benner&Ranganathan,2012). Overall, enticing infomediaries to report favor-ably on a firm by influencing how they (re)constructmeaning regarding the enterprise is a core objective forCEOs (Pfarrer et al., 2010; Rindova & Fombrun, 1999).

We assume that a key element of CEOs’ sense-giving toward infomediaries is “discursive framing”(Cornelissen & Werner, 2014: 183; see also Gioia &Chittipeddi, 1991; Fiss & Zajac, 2006). As suggestedin the social movement and communication litera-ture (Benford & Snow, 2000; Entman, 1993), discur-sive frames are interpretive lenses that actorsstrategically evoke through their communication inorder to shape meaning, mobilize support, and gainlegitimacy (Cornelissen & Werner, 2014: 182). Dis-cursive frames simplify and condense information,and they allow senders of communication toverballyemphasize certain aspects of a given piece of in-formation and suppress others (Clatworthy & Jones,

1 We use the term “journalists” to denote those infome-diaries who cover, among other areas, economic issues onbehalf of the “main street” (Lamin & Zaheer, 2012). Assuch, we focus on generalist (business) journalists andexclude highly specialized infomediaries who report,e.g., in industry-focused media outlets (Petkova et al.,2013).

2 In the following, we use “analysts” and “securitiesanalysts” interchangeably.

1198 AugustAcademy of Management Journal

2003; Entman, 1993; Fiss & Zajac, 2006; Pfarrer et al.,2010). Consequently, given the abundance, com-plexity, and ambiguity of firm-related information,discursive framing should be a central element ofCEOs’ attempts to subtly sway infomediaries to issuemore favorable statements about their firms.

CEOS’ USE OF METAPHORICALCOMMUNICATION: BENEFITS AND COSTS

In this paper,we focus on a particularly important,classical form of discursive framing that has alreadygarnered widespread attention in many researchdisciplines, including management (Cornelissenet al., 2011; Cornelissen & Werner, 2014; Ibañez &Hernandez, 2011; Johnson & Lakoff, 2002; Weick,1998), but has received relatively little attentionin research on executives and their interactionswith stakeholders: metaphorical communication.3

Typically, metaphorical communication maps a fa-miliar and concrete domain of knowledge, knownas the source domain, onto a less familiar andmore abstract domain of knowledge, known as thetarget domain (Lakoff & Johnson, 1980). For in-stance, in the phrase “an organization is a machine”(Heracleous & Jacobs, 2008), the familiar sourcedomain of “machine” is mapped onto the less fa-miliar target domain of “organization.” In so doing,metaphorical communication highlights some char-acteristics of an issue while suppressing others(Entman, 1993; Thibodeau & Boroditsky, 2011)—“Anorganization is a machine” evokes different conceptsof an organization than “an organization is a jazzensemble” (Weick, 1998). As such, metaphoricalcommunication is not merely a rhetorical ornament,but serves as a type of discursive framing becausean issue is virtually “seen through” (Black, 1962: 41)

and conceptualized by the metaphorical expression(Lakoff & Johnson, 1980).

Studies in linguistics and leadership rhetoric haveemphasized the benefits of metaphorical communi-cation for influencing audiences’ sensemaking pro-cesses and actions (Sopory & Dillard, 2002). On acognitive level, metaphorical communication canhelp receivers make sense of information, as it intro-duces something “novel by reference to somethingalready known” (Foster-Pedley, Bond, & Brown,2005: 44), thereby heightening the receiver’s aware-ness, understanding, and retention of a message(Ortony, 1975). Moreover, metaphorical communi-cation facilitates sensegiving, as it allows senders torhetorically distill meaning, and to form and steerstakeholders’ interpretations (Cornelissen & Werner,2014). On an affective level, metaphors link logicaland emotional methods of persuasion by invokingfamiliarity and by referring to sensory experiences(Mio, 1997). Therefore,metaphorical communicationcan engender an overall positive attitude towardcommunicated messages and the senders of thosemessages, and can help align the receiver’s responseswith the sender’s goals (Antonakis et al., 2011; Read,Cesa, Jones, & Collins, 1990).

Notably, this positive description of metaphoricalcommunication is echoed in an abundant stream ofpractitioner-oriented literature, which advises corpo-rate leaders tousemetaphorical communicationwheninteracting with critical audiences (e.g., Antonakiset al., 2012; DenHartog &Verburg, 1997;Miller, 2012;Walz, 2014). Therefore, it is not surprising that CEOsoften use metaphors in their public communication(Amernic, Craig, & Tourish, 2007; Oberlechner &Mayer-Schoenberger, 2002). In 2011, for instance,AOL’s Tim Armstrong used metaphorical communi-cation to announce a new product initiative:

In the Groupon-like coupon business [. . .], we can allexpect a rolling thunder of new products from AOL.(cited in Carlson, 2011; italics added)

However, the scientific management literaturelacks research that has critically examined the out-comes of executives’ usage of metaphorical com-munication in the context of firms’ strategic publiclanguage. This is puzzling because a substantialbody of research has suggested that metaphoricalcommunication might have significant downsides(Merkl-Davies & Koller, 2012; Ramsay, 2004; Steen,2011). In particular, cognitive linguists have pointedout that the familiar source domains used as framesin metaphorical communication are a priori con-ceptual reductions and simplifications (Hamington,

3 Similar to prior studies (e.g., Sopory & Dillard, 2002),we viewmetaphors, similes,metonymies, and analogies aselements of metaphorical communication, as they allcompare “something unfamiliar [. . .] with something fa-miliar” (Corbett & Connors, 1998: 95). While a metaphorimplicitly compares two issues or ideas (“A is B;”e.g., “argument is war”), a simile does so explicitly (“A islike B;” e.g., “employees are like flowers”). A metonymydoes not compare A with B, but substitutes the “A” that isactuallymeantwith a “B” that is attributive or suggestive of“A” (e.g., “The White House said. . .”). Finally, an analogyis an extended comparison in which a causal pattern istransferred from a familiar domain to a less familiar do-main (Corbett &Connors, 1998), as in “Big companies oftenproduce bureaucracy the way gardens produce weeds”(Kindler, 2010: 20).


2009) that involve little detail and precision, neverfully correspond to the target domain (Black, 1962),and never encompass all facets of a given concept(Shenkar et al., 2008). For instance, conceptualizingexploration in organizations through the source do-main of improvisation in a jazz ensemble neglectsthe fact that improvisation in jazz, in contrast to ex-ploration in an organization, typically evolves withina relatively institutionalized structure, including a setof “standard” songs and routinizedmelismas or “licks”(Hatch & Weick, 1998). Moreover, although meta-phorical communication can initially reduce ambigu-ity by focusing receivers’ attention on certain attributesofan issue, it is, bydesign,ambiguousbecause it alwayshas many “potential meanings” (Ramsay, 2004: 146).For example, the “rolling thunder” metaphor used byAOL’s CEOmay include not only positive associationsof power and invincibility but also negative associa-tions of violence and devastation.

When transferred to the CEO–infomediary dyad,these facets of metaphorical communication couldjeopardize processes that would otherwise be ex-pected to sway infomediaries’ assessments posi-tively. On a cognitive level, the lack of precision andthe ambiguity inherent in metaphorical communi-cation could induce the infomediary to frame theinformation conveyed by the CEO differently andless favorably than intended by the CEO. On an af-fective level, metaphorical communication couldcause the receiving infomediaries to develop nega-tive views on the CEO and the firm. At worst, theymight suspect that the CEO is attempting to down-play or even camouflage unfavorable information.Altogether, while there are good reasons to suggestthat CEOs’ use of metaphorical communication af-fects the benevolence of infomediary assessments,the direction of that effect is unclear.

EFFECTS OF CEOS’ USE OF METAPHORICALCOMMUNICATION: A MATTER OF THE

INFOMEDIARY’S CONTEXT

To help resolve these apparent contradictions, wepropose that CEOs’ use of metaphorical communi-cation can lead to favorable or unfavorable ap-praisals, depending on the type of infomediaryaddressed. The key premise underlying our theo-rizing is Wiesenfeld et al.’s (2008: 232) concept ofthe infomediary’s work as “socially situated, [. . .]constituent-minded sensemaking.” This concepthighlights that infomediaries are special in theirsensemaking because, by design, they make sense ofinformation through theprismof giving sense to their

specific audiences. As a result, how infomediariesmake and give sense depends not only on their ownrational analyses and biases, but also on the analysesand biases “they anticipate in their constituents”(Wiesenfeld et al., 2008: 232).

We organize our thinking around two interre-lated assumptions about how infomediaries’ sense-making and sensegiving are socially situated, bothof them rooted in research on social cognition(Fiske & Taylor, 2017). First, we stipulate that info-mediaries, like all humans, are “cognitive misers”who aim tominimize cognitive effort (Taylor, 1981).Thus, we expect infomediaries to interpret infor-mation conveyed by a CEOmore favorably and to bepositively influenced by a discursive frame evokedby a CEO, the more the form of framing is generallyconducive to their own sensegiving (Brown, Call,Clement, & Sharp, 2015; Deephouse & Heugens,2009). Conversely, infomediaries will issue lessfavorable assessments the more a CEO’s framingis inapplicable or even counterproductive to theirwork.

Second, we assume that infomediaries’ cognition—just like social cognition in other (professional)groups—is biased by field-specific, institutionalizednorms and schemas (Bundy & Pfarrer, 2014; Lamin &Zaheer, 2012; Petkova, Rindova, & Gupta, 2013). Spe-cifically,we suppose that infomediaries use the degreetowhich a piece of communication accommodates theidiosyncratic norms and conventions of sensemakingand sensegiving that are shared in their respectivefields as a “cognitive shortcut” (Fanelli & Misangyi,2006: 1053) to judge the quality and trustworthiness ofthe conveyed information and the communicator, andthe overall situation. Thus, if a CEO increases the de-gree to which his or her communication accommo-dates an infomediary group’s institutionalized normsand conventions of sensemaking and sensegiving,members of that group should evaluate the respec-tive firm more favorably. Moreover, infomediariesshould become more skeptical regarding firm-relatedinformation the more they perceive the CEO’s com-munication as violating their professional norms andconventions.

Building on the above premises, we assume thatinfomediaries’ appreciation of metaphorical com-munication is socially situated and that it differsamong various types of infomediaries because suchrhetoric is likely to suit the institutionalized needsand norms of some groups of infomediaries but notthose of others. In the following, we further developand formalize this rationale using the examples oftwo highly important, profoundly dissimilar, and


frequently studied types of infomediaries: journal-ists and securities analysts.

CEOs’ Use of Metaphorical Communication andJournalists’ Reporting

Journalists are one of the most important infome-diary groups (Deephouse, 2000), and preferred ad-dressees of CEO communication (Westphal et al.,2012), because they operate at the interface betweenthe firm and broader society (Gamson & Modigliani,1989). We propose that journalists’ sensemakingand sensegiving are socially situated in a way thatrenders journalists favorable toward metaphoricalcommunication.Asdescribed in sociological studiesof journalism, journalists aim to enlighten the publicin a way that goes beyond the mere reproduction ofinformation (Shoemaker & Reese, 1996). They wantto raise interest in their reporting and give “legiti-macy and credibility to what they do” (Deuze, 2005:446). To achieve these goals, journalists must pro-vide broad, easy-to-grasp, and engaging informationabout the firms they cover (Andsager, 2000; Tuchman,1972). As suggested by the media-dependency hy-pothesis (Ball-Rokeach & DeFleur, 1976), journalistscan give suchmeaning not only by relaying digestiblepieces of important information (Deuze, 2005) butalso by providing frames that fit the public’s reality(Gamson & Modigliani, 1989) and “resonate with[its] existing underlying schemas” (Scheufele &Tewksbury, 2007: 12). Therefore, metaphorical com-munication isparticularly suited for journalisticworkbecause it allows journalists to forgo complex, tech-nical explanations, and instead build on their audi-ences’ experiences and schemas (Lakoff, 1993).Givenour assumptions, then, we expect journalists to re-spond positively if a CEO increases his or her use ofmetaphorical communication because translatingmetaphorically framed content into their own sense-giving requires them to expend less cognitive effortthan translating literally communicated content.

In light of the communicative needs of journalists’audiences, it is not surprising that metaphoricalcommunication has long been part of their rhetoricalcanon—their “thought world” (Lamin & Zaheer,2012: 47) and socialization, and their training. No-tably, teachers of journalism often advocate for theuse of metaphorical communication, even in busi-ness journalism (e.g., Burns, 2013;Morley, 2007). Forinstance, Peter Coy, economics editor of BloombergBusinessweek, argued that “[t]rying to communicatewithout using any metaphors would be like trying tocomplete a paint-by-numbers canvas without red,

blue, yellow and green” (2013). Metaphorical com-munication also resonateswith the social backgroundof most journalists, most of whom have degrees indiscursive and text-focused disciplines, and workwith words rather than numbers (Medsger, 2014).4

Correspondingly, scholars have long observed the mul-tiplicity and variety of metaphors used by journalists,especially in business journalism (Partington, 1995). Insummary, metaphorical communication is part of jour-nalists’ institutionalized norms and schemas, and jour-nalistsviewthis typeofcommunicationasan indicationofquality,eloquence,andcompetence.Assuch, ifaCEOuses more metaphorical communication, he or she bet-ter accommodates journalists’ institutionalized normsof sensemaking and sensegiving, ultimately leading tomore positive assessments from journalists.

Hypothesis 1. Ceteris paribus, the more a CEO usesmetaphorical communication, the more favorablyjournalists will report about that CEO’s firm.

CEOs’ Use of Metaphorical Communication andAnalysts’ Evaluations

Analysts are focal addressees of CEO communica-tion because they gather and interpret market- andfirm-specific information to issue research reports forinvestors, which include earnings forecasts and ad-vice on whether to buy, hold, or sell stocks (Giorgi &Weber, 2015). We expect analysts, in contrast tojournalists, to respond negatively if a CEO increaseshis or her use ofmetaphorical communication. This isbecause, first, the success and status of analysts de-pend onwhether they provide detailed, accurate, andclear recommendations and reports (Giorgi & Weber,2015). As noted above, metaphorical communicationis limited in detail, rather inaccurate, and inherentlyambiguous (Ramsay, 2004). Therefore, it is inappli-cable to analysts’ work for the same reasons that it isapplicable to journalists’work. In particular, themorea CEO uses metaphorical communication, the morecognitive effort an analyst must expend to contextu-alize, interpret, and ultimately translate the informa-tion intoprecise recommendations.Analysts also findit challenging to juxtapose metaphorical CEO com-munication with their own insights and forecasts.They may therefore perceive metaphorical commu-nication as distracting “noise.”All of these factors are

4 We are grateful to leading journalists and professors of(business) journalismwhomwe interviewed as part of thisstudy. They confirmed our understanding of (business)journalists’ social and educational background.


likely to bias analysts negatively and to reduce theodds that theywill adopt the CEO’s interpretations intheir own sensegiving.

Second, given the requirements of the analyst’sprofession, metaphorical communication is notengrained in their rhetorical canon.Analysts usuallyhold degrees in computational disciplines, such asfinance, economics, and accounting, or in computerscience, physics, or engineering (Block, 1999).Manyanalysts have MBAs and are certified as CharteredFinancial Analysts (Block, 1999; Brown et al., 2015).In contrast to journalists, analysts develop detailedpresentations and financial reports, primarily byusing “spreadsheets, relational databases and sta-tistical and graphics packages” (Granville, 2014: 1).Thus, while journalists appreciate the familiarityand generalness of metaphorical frames, analystsoperate in a thought world that is structured bynumbers and facts (Fuller & Metcalf, 1978), which isat odds with a metaphorical representation of re-ality.5 Combining these insights with our premisethat infomediaries inherently respond unfavorablyto communication that is incongruent with theirfield-specific professional norms and schemas, weconclude that analysts will generally respond skep-tically the more a CEO uses metaphorical commu-nication. We therefore hypothesize:

Hypothesis 2. Ceteris paribus, the more a CEO usesmetaphorical communication, the more unfavorablyanalysts will evaluate that CEO’s firm.

The Moderating Effect of NegativeEarnings Surprises

Our theorizing is based on the notion that infome-diaries use certain aspects of firms’ public language—inour case, CEOs’ use of metaphorical communication—as a “cognitive shortcut” in their appraisals (Fanelli &Misangyi, 2006: 1053). Part of whatmakes these biasedinterpretations so intriguing is that, according to bothgeneral and capital-market-specific social cognitiontheory (Fiske & Taylor, 2017; Gao et al., 2015; Healy &Palepu, 2001), human reliance on cognitive shortcutsvaries, depending on other facets of the providedinformation.

We argue that, in the context of infomediaries’assessments of CEOs’ public communication, infor-mation about firm performance relative to expecta-tions will be a particularly influential moderator.Performance that positively or negatively deviatesfrom market expectations—so-called “earnings sur-prises” (e.g., Brown, 2001)—is especially importantinformation from the perspective of most infome-diaries, as such deviations might require them toreassess the firm and its future (Pfarrer et al., 2010;Washburn & Bromiley, 2014). Moreover, a signifi-cant body of cognitive psychology has suggestedthat whether a given piece of information is positiveor negative strongly influences human behavior,including human reliance on cognitive shortcuts(Rozin & Royzman, 2001; Taylor, 1991). More pre-cisely, research on the “negativity bias” has suggestedthat people who act in situations of uncertainty,ambiguity, and pressure tend to rely more on cogni-tive shortcuts when they interpret negative andpessimistic information than when they interpretpositive and optimistic information (Baumeisteret al., 2001).6

5 We checked whether the supposed (dis-)inclinationtoward metaphorical communication is really reflected injournalists’ and analysts’work. For two of the firms in oursample, we randomly selected 100 articles from the NewYork Times and Wall Street Journal that mentioned therespective firms at least once in the text, as well as 100analyst reports covering these firms. From the newspaperarticles and analyst reports, we extracted each statementcontaining the name of at least one of the respective firms.We then followed the coding guideline for metaphoricalcommunication that we present in this paper and checkedwhether any metaphorical communication appeared inthose statements. 8.7% (27 out of 309) of journalists’statements employed metaphorical communication whenreferring to these firms. In contrast, only 0.08%of analysts’statements (7 out of 8,243) used metaphorical communi-cationwhen commenting on the focal firm. In linewith ourtheorizing, journalists usedmetaphors such as “$5million[. . .] would look like bus fare to the four big players in thestent business” or “Amgen Inc. and Johnson & Johnsonhave taken their long-running blood feud to Capitol Hill,”which are creative and clearly nonidiomatic (see themethod section for our precise definition of metaphoricalcommunication as compared to idiomatic language).Conversely, the few metaphors used by analysts wererather common and nearly idiomatic, such as “Johnson &Johnson needs to overcome several roadblocks.”

6 This effect most likely unfolds for evolutionary rea-sons. Generally, it is evolutionarily useful to give negativeinformation more consideration than positive information(Baumeister et al., 2001). Moreover, when an event occursand individuals lack comprehensive information on howto adequately respond to it, they typically interpret theevent by relying on their engrained and “tried and tested”cognitive schemas (Taylor, 1991). As a result, assessmentsof negative events will be more biased by cognitive short-cuts than assessments of positive events will (assuminga situation of evaluative uncertainty and ambiguity).


In this vein—and considering the considerableevaluative uncertainty and the omnipresent timepressure under which infomediaries work (Fanelliet al., 2009; Tuchman, 1972)—we argue that themorethat a firm’s performance negatively surprises, themore infomediaries’ assessments will be biased bytheir engrained schema regarding CEOs’ use of meta-phorical communication. For the case of journalists,given their favorable schema of CEOs’ use of meta-phorical communication, this implies that the in-crease in journalists’ favorability that stems from anincrease in the CEO’s use of metaphorical communi-cationwill be greater themore the firm’s performancedisappoints. In particular, we envision an increaseduse of metaphorical communication by a CEO tosubtly indicate to the journalist that the CEO is par-ticularly ready and capable of dealing with the situa-tion. For the case of analysts, the interactive effect ofmetaphorical communication and negative earningssurprises on analysts’ favorabilitywill be the oppositebecause, as suggested in Hypothesis 2, analysts viewmetaphorical communication with skepticism oreven as an attempt to camouflage unpleasant facts.

Hypothesis 3a. Ceteris paribus, the more negative thefirm’s earnings surprises, the stronger the positivemarginal effect of the CEO’s metaphorical commu-nication on the favorability of journalists’ reportingabout the firm.

Hypothesis 3b. Ceteris paribus, the more negative thefirm’s earnings surprises, the stronger the negative mar-ginal effect of the CEO’s metaphorical communicationon the favorability of analysts’ evaluations of the firm.

METHODS

Given the systematic differences in data on jour-nalists’ and analysts’ evaluations of firms, and theo-retically motivated differences in the set of controls,we ran two analyses. Analysis I estimates journalists’evaluations based on their statements about firms,whileAnalysis II estimatesanalysts’evaluationsbasedon their (a)EPS forecasts and (b) recommendations.Toallow for a comparison of Analysis I and Analysis II,weuse the samesampleof firmsand the samemeasureof CEOs’ use of metaphorical communication.

Sample

We drew our sample from the population of firmsthat operated in the U.S. pharmaceutical industryand the U.S. computer hardware and software in-dustries between January 1, 2002, and July 31, 2011.

These industries are suitable for our study becausethey are characterized by a high level of CEO dis-cretion (Hambrick & Finkelstein, 1987), and are thesubject of abundant coverage by journalists and an-alysts. We collected data from the Osiris, Mergent,Ward’s Business Directory, and Thomson SDC Plati-num databases to identify all firms that met the fol-lowing criteria: (1) a Global Industry ClassificationStandard (GICS) of 4510 or 4520 in the hardware andsoftware industries, or a GICS of 3510 or 3520 in thepharmaceutical industry;7 (2) headquartered in theU.S. in2002; (3) revenueofmore thanUSD100millionin 2002; and (4) two or more CEOs in the period be-tween 2002 and 2011. Criteria 1 to 3 were chosen inorder to obtain a sample of comparable firms thatwere sufficiently large to receive substantial coveragefrom journalists and analysts. Nevertheless, we ex-cluded 22 firms owing to the limited availability ofconference-call transcripts, the discursive vehicle weused to capture CEOs’ metaphorical communication.Criterion 4 allowed us to better discern effects at theCEO and firm levels. Subsequently, we excluded 15extraordinary cases involving certain types of CEOs,particularly interim CEOs or co-CEOs, which mighthave confounded our analysis by, for instance, in-troducing a sampling bias toward poor performers(Krieger & Ang, 2013). After further reductions due tomissing data for the controls,8 our final samples con-sisted of 43 firms for Analysis I (n5 449 comparisonsbefore or after the conference call; 98 CEOs) and 47firms for Analysis II (n 5 624 comparisons of aggre-gated analyst forecasts before or after the conferencecall, 101CEOs; 270 comparisons of aggregated analystrecommendations, 94 CEOs).9

7 In the pharmaceutical industry, we excluded firms thatderived less than 40% of their sales from pharmaceuticals(following the approach of Kaplan, Murray, & Henderson,2003).

8 We tested for sample attrition and sample selection(Wooldridge, 2010) by constructing a sample-selection in-dicator, sit, which specifiedwhether we observed all xit andyit. (Notably, in our main analysis, we do not use observa-tionswhen sit5 0 because data for at least some elements of(xit yit) are unobserved in these cases). Fixed effects are in-consistent if the sample selection is not strictly exogenous.Therefore, the selection indicator from other time periods(e.g., sit 1 1) should be insignificant at time t. We calculatedthe robust t statistic for sit 1 1 in yit 5 xitb1 uit 1 sit 1 1 andfound no significant effect for this selection indicator.Therefore,we conclude that themissing observations in ourpanel do not follow a systematic pattern.

9 We reran all analyses with the overlap of the samples.The results remained consistent.


Independent Variable: CEOs’ Use ofMetaphorical Communication

Discursive vehicle. We selected firms’ quarterlyearnings conference calls as the focal discursivevehicle of our study for three reasons. First, participa-tion in conference calls is a vital source of firm-relatedinsights forbothanalysts and journalists (Bushee,Core,Guay, & Hamm, 2010; Jorgensen & Wingender, 2004;Roush, 2011).10 Second, both infomediary groups aresimultaneously and directly affected by CEO rhetoricduring conference calls. Third, the use of conferencecalls ensures comparability between firms, and allowsus to inherently control for many other potential in-fluences on the relationship between CEOs’ rhetoricand infomediaries’ evaluations. For example, confer-ence calls take place in similar settings, have a rela-tively standardized length, and cover similar topicsacross firms.We usedThomsonResearch and SeekingAlpha to obtain transcripts of quarterly earnings con-ference calls.Our final sample covered937hand-codedconference calls, which constituted approximately8,000 pages of text.

Coding process. Guided by Mio et al.’s (2005)approach, we iteratively developed a reliable, context-sensitive, noncomputerized content-analytical in-strument to identify and measure CEOs’ use ofmetaphorical communication (Krippendorff, 2004).The coding process had three phases. In the firstphase,wedevelopedpreliminary coding instructions,including concise definitions of the rhetorical ingredi-ents of metaphorical communication (i.e., metaphors,similes, analogies, and metonymies). We providedanchoring examples, coding criteria, and intersubjec-tively comparable guidelines that illustrated how toidentify metaphorical communication reliably (Miles& Huberman, 1994).

In the second phase, two of the authors and threespecially trained coders independently pretested thecoding instructions by hand-coding 25 transcripts ofconference calls held by U.S. pharmaceutical com-panies not included in the final sample. Together,these actors discussed inconsistent codings untilthey arrived at an agreement (Krippendorff, 2004).Thereafter, we revised the initial coding instructionsand supplemented them with various examples ofCEOs’ metaphorical communication. We also de-cided to exclusively focus on what we termed“contentual”CEO communication; i.e., we excludedpassages in which the CEO welcomed participants,

exchanged compliments, and directed questions toother firm representatives.11

In the third phase, we applied the initial codingguidelines to the conference calls in the sample andoptimized our coding instruments. More specifi-cally, three two-person teams of trained coders in-dependently coded all conference-call transcripts.The teams then met to compare and discuss everyidentified metaphor. As part of this process, we alsospecified whether metaphors were “dead meta-phors” and, therefore, had to be excluded from thecoding. “Dead metaphors” are metaphors that have“become so familiar and so habitual that we haveceased to be aware of their metaphorical nature anduse them as literal terms” (Tsoukas, 1991: 568), suchas “on the one hand . . . on the other hand.”12

Throughout this phase, we gauged the robustnessof our coding. First, we tested interrater agreementwith satisfactory results (Krippendorff’s [2004] a 50.74). Two types of interrater disagreement werecommon: one in which the codings deviated withregard to how many words should be counted asbelonging to a specific metaphorical expression, andanother in which metaphorical communication wasonly recognized by some of the coders. In cases ofcontinued disagreement, the first author acted as anindependent evaluator and made a final decision.Subsequently, the teams created a final version ofeach coded document.13

10 Notably, journalists regularly refer directly to theseconference calls in their reporting.

11 Examples of “noncontentual” communication are “. . .good question [!];” “Thanks for participating in all the goodquestions, and we look forward to seeing you at . . .” (bothMike Fister, Q3 2006); and “Mary Kay, why don’t you gointo the details of this, make sure I don’t misspeak” (RobertParkinson, Q4 2009). We excluded these statements be-cause we theorize about CEOs’ sensegiving regarding thefirm, which is not the topic of noncontentual communi-cation. Note that we use noncontentual communication asa control in our analyses.

12 We followed prior linguistics research (PragglejazGroup, 2007) in classifying a metaphor as dead if it hasbecome so conventionalized that its meaning is explainedin an ordinary dictionary. We referred to the Merriam-Webster and Cambridge Dictionaries. We also treatedidioms as dead metaphors, as they have ceased to be figu-rative and their meaning has become routine (Burbules,Schraw, & Trathen, 1989). Technical jargon, such as theterm “pipeline” in the pharmaceutical industry, wastreated in the same way (Lindsley, 1991).

13 Complete coding guidelines, including the list of deadmetaphors, and an extended list of examples of meta-phorical communication used by the CEOs in our samplecan be obtained from the authors.


Table 1 provides examples of metaphorical com-munication used by the CEOs in our sample. In total,we identified 2,229 instances of metaphorical com-munication in our final sample of conference calls,the majority of which (95%) took the form of meta-phors. As can be expected based on prior work (Mioet al., 2005), the CEOs in our sample used a widevariety of metaphorical communication to framea broad range of topics, including their firm’s per-formance outlook, business partnerships, and productpolicies.

Measure of CEOs’ use ofmetaphorical communi-cation.Ourgoalwas todevelopameasure that reflectedthe weight of metaphorical communication as part ofthe overall length of the communication. To do so, wefirst counted allwords belonging to a coherent sentencestructure (i.e., subject, predicate, object) that werenecessary to make sense of a given metaphoricalexpression, and classified those words as meta-phorical communication. We then operationalizedCEOs’ use of metaphorical communication by di-viding the total number of words in a CEO’s

TABLE 1Examples of Metaphorical Communication Used by CEOs in Conference Calls

CEO Statement Type of Figure CEO (Company/Year/Quarter) Target Domain

“We didn’t tell you what it was, and you will have towait until we announce it. Just like Christmas. ButSanta is coming.”

Analogy Michael Dell (Dell/2004/Q3) Product launch

“. . .it’s a little bit like breaking in a new Maserati. Thefirst thousand miles, you are not going to step onthe gas too hard.”

Analogy Howard Pien (Chiron/2005/Q3) Restructuring process

“Both parties have input on a plan, a detailed plan, soI would say we are both in the front seat of the car.In Phase I, we are in the driver’s seat; in Phase II,they take over the driver’s seat. But each is navigatingwith the other.”

Metaphor Daniel Welch(InterMune/2007/Q1)

Business partnership

“It feels like we just finished the preseason and we’resuited up now and ready to play the Super Bowlagain this year.”

Metaphor Brad Smith (Intuit/2007/Q1) Performance outlook

“We’ve been actually watching that fairly closelybecause otherwise you build a kind of a ticking timebomb, and certainly we don’t want to do that [. . .].”

Metaphor Norman Schwartz(Bio-Rad/2008/Q4)

Inventory level

“It is not a fixed panel or closed system. I like to thinkabout it just like the iTunes music model, wherecustomers can pick and choose their own play listsand [are] not necessarily constrained to buying anentire album when all they want to purchase is asubset of the information.”

Analogy Kevin King (Affymetrix/2009/Q3) Product policy

“We are lean, but we have, I would say, good muscles.We are in good shape. We should be able to runpretty fast whenever it’s required.”

Metaphor Lukas Braunschweiler(Dionex/2009/Q3)

Performance outlook

“So look, at the end of the day, our customers want acheaper price, we want a higher price, so the battlewill be fought on that basis [. . .].”

Metaphor Steve Dubin (Martek/2009/Q4) Pricing

“That’s the biggest dark cloud that we’re continuallylooking at, and then the sunshine that’s lurkingbehind that is the commercial refresh and the rate atwhich that progresses. There is a moon there as wellwhich is the strength in Asia, which is significant.”

Analogy John Coyne (WesternDigital/2010/Q4)

Performance outlook

“And then, on the front of that, Todd, if you thinkabout Bayesian-type forecasting algorithms, whichis how they forecast hurricanes. Being from SouthLouisiana, I know all about that. You watch everyday and see how it moved and then how—whereit’s expected to strike landfall, and the Clinical trialis the same way.”

Analogy Joseph Herring(Covance/2011/Q1)

R&D

Note: Data shown in chronological order.


metaphorical communication during a conferencecall by the total number of words in the CEO’scontentual communication during the same call.14

Moderating Variable: Negative Earnings Surprises

To gauge the degree to which firm performancewas below market expectations, we first computeddeviations from market expectations (e.g., Brown,2001; Pfarrer et al., 2010). Specifically,we calculatedthe difference between a firm’s quarterly EPS and themean of analysts’ EPS forecasts for that quarter,scaled by the actual EPS. In line with our theoreticalarguments, we splined this variable and includednegative earnings surprises (i.e., earnings below themean of analysts’ EPS forecasts) and, as a check,positive earnings surprises (i.e., earnings above themean of analysts’ EPS forecasts) as moderators.15

Dependent Variable in Analysis I: Favorability ofJournalists’ Reporting

Our measure of the favorability of journalists’reporting largely follows approaches found in priorresearch (Deephouse, 2000; Pollock &Rindova, 2003).We gauged how journalists’ assessments of firmschanged from (a) the period between the prior con-ference call and the focal conference call to (b) theperiod between the focal conference call and the fol-lowing conference call.16 We conducted a manualcontent analysis of journalists’ statements about eachfirm in our sample published in the New York Times(NYT) and theWallStreet Journal (WSJ),which are thetop-circulating national newspapers in the UnitedStates (Wolfe, 2012). We chose to focus on theseleading outlets instead of randomly selecting state-ments from a broad range of newspapers as doing soallowed us to avoid the bias that stems from mimetic

“pack journalism” (Williams, 2011). Moreover, theexperts we interviewed emphasized that journalistswriting for the NYT and theWSJ use conference callsparticularly intensely in their reporting.

To collect a meaningful and manageable amountof data, we first searched Factiva for all articlesthat appeared in either of the two newspapers be-tween 2001 and2011 andmentioned the firm’s nameat least once. We then followed the progressivearticle-selection process developed by Deephouse(2000).17 This sampling procedure yielded a total of10,155 articles. We then extracted and read eachstatement that contained the name of the respectivefirm. We also carefully read the sentence that fol-lowed the focal statement in order to extract state-ments with indirect but unambiguous mentions ofthe focal firm, such as “the company” or “the soft-ware maker.” Moreover, to ensure that we onlycaptured journalists’ favorability (and not analysts’evaluations), we excluded 947 statements in whichjournalists directly quoted analysts. This procedureyielded a total of 25,415 statements.

Two trained coders then worked with two of theauthors to develop a comprehensive coding protocol(Miles & Huberman, 1994) to reliably group thestatements into three categories. The first categoryincluded statements that favorably portrayed thefocal firm. The second contained statements thatwere ambiguous (i.e., contained both positive andnegative evaluations, or messages that could beinterpreted both positively and negatively). Thethird category included unfavorable statementsabout the focal firm. We first had multiple coders,including the authors, collectively assess 150 state-ments, after which we relied on independent codersto code the remainder of the statements.A systematictest (Lacy & Riffe, 1996) showed highly acceptableinterrater reliability (Krippendorff’s [2004]a50.83).Nevertheless, throughout the process, the coders andthe authors discussed unclear cases to ensure con-sistent and reliable coding.

Similar to prior studies (e.g., Pollock & Rindova,2003), we calculated the Janis–Fadner coefficient ofimbalance (Deephouse,2000; Janis&Fadner, 1965) tomeasure the favorability of journalists’ reportingabout a firm.Given thatwe theorize about journalists’assessments of specific firms, we considered the

14 Other operationalizations, such as counting instancesin whichmetaphorical expressions are used by CEOs (Mioet al., 2005) or the absolute number of the CEO’s meta-phorical words in a given conference call, yielded resultsthat were consistent with those we report here.

15 We found more positive values and a higher mean forthe positive earnings surprise spline. This is in line withthe argument that firms actively attempt to avoid negativesurprises.

16 To reduce the likelihood that confounding events bi-ased our results, we ran robustness checks with recalcu-lated measures of journalists’ favorability (i.e., includingstatements appearing only in the 60, 30, and 20 days fol-lowing the focal conference call). The results were con-sistent with those reported here.

17 In addition to Deephouse’s (2000) process, for firmsthat yielded more than 24 articles in a given year, we ran-domly selected 24 articles (the average number of articlesper firm per year in our sample). The complete selectionand coding guidelines can be obtained from the authors.


individual statement as our recording unit (Deephouse,2000), and calculated the ratio of favorable statementsto unfavorable statements in the period beginningright after a focal conference call and ending just be-fore the next conference call, while controlling for thetotal number of statements in that period (i.e., includ-ing statements that were neutral). In so doing, weensured considerable temporal proximity between aCEO’s communication and a journalist’s assessment.Finally, as the “measurement of a dependent variable attwopoints in time iswidely regardedapowerful tool formaking causal inferences from nonexperimental data”(Allison, 1990: 93), we computed the d of the favor-ability after the focal conference call to the favorabilityin the period before the focal call (the latter defined asthe time between the prior conference call and the focalconference call).18

Control Variables in Analysis I

We included the following control variables inAnalysis I. Table A2 in Appendix A summarizes thedata sources for all variables.

Firm controls. We used firm fixed effects estima-tors toaccount fordifficult-to-observedifferencesamongfirms that are invariant over time. We accounted foradditional firm-level explanations by including the fol-lowing time-variant factors that, as they change, couldsignificantly affect the favorability of journalists’ report-ing (Westphal & Deephouse, 2011). Prior firm perfor-mance change was operationalized as the change inreturn on assets (ROA; measured as the ratio of net in-come to total assets) in the 12 quarters preceding thefocal quarter. Prior firm performance volatility wasmeasured as the standard deviation of the ROA in thesame12quarters (Fanelli et al., 2009).Aschanges in firmsize might affect whether a company is subjected tomedia critique (Fang & Peress, 2009), we also includedthe natural logarithm of total assets at the end of thequarter preceding the focal conference call. Further-more, we included the number of press releases issuedby the focal firm in a given year, as press releases areintended to affect the scope and tone of reporting byjournalists (Pollock & Rindova, 2003). Similar to priorstudies (Chatterjee &Hambrick, 2011),we also includedmediaattention to the firmbycounting the total numberof statements in the NYT and WSJ that mentioned thefocal firm in the period after a conference call.

CEO controls. We included a set of variables thatcould affect how a CEO and his or her firm are per-ceived by infomediaries. In this regard, we controlledforCEOage andCEO tenure.Moreover, we gauged theCEO’s structural power (Finkelstein, 1992) usinga dummy for cases in which the CEO was also chair-man of the board (CEO duality; coded as 1). We alsocontrolled for an incoming CEO’s status as contender,outsider, or follower, which could affect journalists’appraisals around the time of a succession (Shen &Cannella, 2002). A CEO was coded as a contender(coded as 1) if he or she was an insider successorreplacing a CEOwhose time in office ended before theage of 64. ACEOwas coded as an outsider (coded as 1)if he or she was not previously employed by the firm.We classified all other CEOs as followers but omittedthis category in our analysis. In line with Fanelli et al.(2009), we alsomeasured overallmedia attention paidto the CEO by counting how often a CEO was men-tioned in theNYT and theWSJ in the quarter before thefocal conference call. Relatedly, we measured CEOcelebrity (Wade, Porac, Pollock, & Graffin, 2006) bycounting the number of “American Business Awards”and “International Business Awards” a CEO receivedin a given year (Chatterjee & Hambrick, 2011).19 Fi-nally, we controlled for CEOs’ functional backgrounds(Ocasio & Kim, 1999) using dummy variables for theCEO’s prior experience in sales and marketing andtheCEO’spriorexperience in finance, asbothmayaffectCEOs’ communication and infomediaries’ appraisals.

Journalist controls. To account for systematic dif-ferences between newspapers (Deephouse & Heugens,2009), we controlled for how many of the statementsissued in the period after the conference call werepublished in the NYT and the WSJ, respectively. Welabeled this variable number of statements.

Conference-call controls. To control for the factthat a CEO’s relative involvement in a conferencecall could influence journalists’ favorability, we in-cluded the share of the CEO’s words of all wordsspoken by firm representatives during the confer-ence call. Moreover, given the importance of thechief financial officer (CFO), especially for providingdetailed information (Larcker & Zakolyukina, 2012),

18 We ran robustness checks to account for the fact that,under some conditions, change scores may lead to in-accurate findings (Allison, 1990). See Appendix A for ad-ditional details.

19 These awards are categories of the Stevie Awards (seewww.stevieawards.com). We used a yearly measure as theawarding procedure evolves over a prolonged period everyyear: Finalists are announced in May, and the awardsbanquets take place in June and September. Winners areprominently featured on the website for the rest of the year.We expect all of these events to draw public attention and,in combination, add to a CEO’s celebrity in the focal year.


http://www.stevieawards.com

we controlled for CFO involvement, which we mea-sured as the ratio of the words spoken by the CFO tothe words spoken by the CEO.

In addition, we controlled for six aspects of CEOcommunication during the conference calls. First, fu-ture orientation (Matsumoto, Pronk, &Roelofsen, 2011)wasmeasured as the ratio of future-oriented to present-oriented words used by the CEO. We identified thesewordsusing the“future”and“present”categoriesof theLinguistic Inquiry Word Count (LIWC) dictionary(Pennebaker, Booth, & Francis, 2007). Second, we ap-plied the “primary process” subcategory ofWordStat’sRegressive Imagery Dictionary (Martindale, 1990) tocapture the share of image-based language, which usescertain words to create a sensory experience (Black,1962) and has been described as influencing thefavorability of receivers (Carton, Murphy, & Clark,2014). Third, based on the assumption that infome-diaries might welcome easy-to-understand communi-cation, we controlled for the comprehensibility of CEOcommunication, which we measured using the Gun-ning Fog Index (Li, 2008). Fourth, given that part of ourlogic is related to the informational needs of journalistsand analysts, and assuming that CEOs can use meta-phorical communication while simultaneously pro-viding detailed facts and figures in other parts of thecommunication (Henry, 2008), we captured the CEO’sfact orientation by measuring the CEO’s use of numer-ical language as a proxy. To do so, we counted howoften a CEO used the terms included in the LIWC“numbers” category (Pennebaker et al., 2007) and di-vided that sum by the number of CEO’s contentualwords. Fifth, as infomediaries might be swayed byCEOs’ compliments and acknowledgments (Westphal& Deephouse, 2011), we controlled for the CEOs’ non-contentual communication by counting the ratio of theCEO’s noncontentual words in a call to all words spo-kenby theCEO.Sixth,we included theoptimismof theCEO’s tone (e.g., Guo, 2014) by employing Loughranand McDonald’s (2011) dictionary to detect positiveand negative words used by the CEO and then calcu-lating the Janis–Fadner coefficient of imbalance(Deephouse, 2000; Janis & Fadner, 1965) for each con-ference call.

Finally, to control for additional time-specific effectsnot captured in our controls, we included dummyvariables for all quarters covered in the sample.

Econometric Approach in Analysis I

We applied a robust firm fixed effects panel esti-mator using the Huber–White standard error correc-tion (xtreg, fe robust in Stata), as a Wald test indicated

heteroskedasticity in our fixed effects model.20 In allinteraction tests, we mean-centered the componentvariables.

Dependent Variable in Analysis II: Favorability ofAnalysts’ Evaluations

Analogous to Analysis I, we used the change in an-alysts’ favorability from the period before the focalconference call to the period after the call as our de-pendent variable. As prior work has used both EPSforecasts and analyst recommendations to gauge ana-lysts’ assessments—with equally good rationales—wedeveloped two different but complementary measuresandused them in separate estimations. The firstmeasurefocusesonthechangeinanalysts’EPSforecasts (Francis&Soffer, 1997). We searched the Thomson Reuters In-stitutional Brokers Estimates System (I/B/E/S) detailedforecast database to collect all EPS forecasts issued by theanalysts covering the firms in our sample. To lower theprobability of alternative explanations, we only includedEPS forecasts from analysts who had issued a forecast intheperiodbetween theprior conferencecall and the focalcall (“last EPS forecast before call” in the formula below)anda forecast in theperiodbetween the focal call and thenext call (“first EPS forecast after call”).21 Moreover, toensure comparability, we only considered one-year EPSforecasts. This resulted in a sample of 6,969 before–afterpairs of EPS forecasts. We calculated the difference be-tween an analyst’s first EPS forecast after the call and thatanalyst’slastEPSforecastbeforethecall.Westandardizedthismeasure by the share price at the close of the quarterforwhich the focalconferencecallwasheld (Fanelli et al.,2009). Tomake this analysis more comparable to Analy-sis I,we thenaggregated the analysts’ forecasts ona grouplevel by calculating the median change in favorabilityacross all analysts for the specific call and firm.22 Finally,

20 Aswehadmissingdata,wewonderedwhether itwouldbe more appropriate to use pooled ordinary least squares(OLS) estimation. However, for panel data, pooled OLS be-comes biased when there is evidence of a fixed firm effect.Accordingly,weconductedanF-test to evaluatewhether thefixedeffectsuiareequal tozero inourmodel (yit5xitb1ui1eit).We foundstrong evidencepointing to aneed to reject thenull hypothesis that the fixed effects are zero (p , 0.001).Therefore, we are required to estimate fixed effects.

21 If the same analyst issuedmore than one EPS forecast/recommendation before or after the focal conference call,we chose the evaluation that was issued closest to the call.

22 As we describe in Appendix A, we also took the op-portunity to test the effect of CEOs’ use of metaphoricalcommunication at the individual analyst level.


we multiplied by 100 to ensure more interpretable re-gression coefficients.

Our secondmeasure of analysts’ favorability focusedonanalyst recommendations (Fanelli etal., 2009).Tobeconsistently conservative across our approaches, weidentified those analysts issuing at least one recom-mendation in the period between the prior conferencecall and the focal call and one recommendation in theperiod between the focal call and the following call.23

Thisyielded393before–afterpairsof recommendationsfrom individual analysts. Furthermore, to maximizethecomparabilitybetween this examinationof analysts’favorability and our examination of journalists’

favorability, we first classified recommendations aspositive (I/B/E/S code1 and2), negative (I/B/E/S code 4and 5), or neutral (I/B/E/S code 3), and then calculatedthe Janis–Fadner coefficient before and after the call.24

The final recommendation-based measure was the dif-ference between the two scores.

Control Variables in Analysis II

Similar to Analysis I, we included a dummy forevery quarter from 2002 to 2011. In addition, for bothsets of estimations—the one estimating changes inEPS forecasts and the one estimating changes in theJanis–Fadner index of analysts’ recommendations—we followed prior research on securities analysts bycontrolling for several factors.

Firm controls. As in Analysis I, Analysis II useda firm fixed effects estimator, and included prior firmperformance change, prior firm performance volatility,firm size, the number of press releases, and media at-tention to the firm.Tocontrol for changes inother time-variant firm-level indicators that are typically assumedto affect analysts’ favorability (Chen&Cheng, 2006),we

TABLE 2Correlations and Descriptive Statistics of Analysis I

Variable Mean SD 1 2 3 4 5 6 7 8 9 10 11

1 Favorability of journalists’ reporting 0.03 0.32 1.002 CEO’s use of metaphorical communication 0.01 0.01 0.00 1.003 Prior firm performance change 0.01 0.12 20.06 0.00 1.004 Prior firm performance volatility 0.09 0.09 20.04 20.02 0.13* 1.005 Negative earnings surprise 0.05 0.15 0.03 20.09* 20.01 0.06 1.006 Positive earnings surprise 0.16 0.29 0.07 0.05 0.11* 0.16* 20.17* 1.007 Firm size 8.66 1.65 20.10* 0.12* 20.19* 20.35* 20.30* 20.22* 1.008 Number of press releases 441.34 556.05 20.05 0.01 20.22* 20.25* 20.15* 20.10* 0.63* 1.009 Media attention to the firm 149.28 241.77 20.06 0.07 20.18* 20.22* 20.11* 20.09 0.57* 0.87* 1.00

10 CEO age 52.74 6.51 0.00 0.05 0.03 20.03 20.07 20.15* 0.26* 20.07 20.03 1.0011 CEO tenure 3.61 3.60 0.00 20.06 0.09 0.05 0.01 0.06 20.15* 20.06 20.08 0.10* 1.0012 CEO duality 0.50 0.50 20.03 20.08 20.02 0.01 20.16* 20.03 0.21* 0.14* 0.15* 0.21* 0.29*13 Contender 0.49 0.50 0.00 20.02 0.01 0.15* 0.03 0.03 20.12* 20.20* 20.22* 0.04 20.26*14 Outsider 0.34 0.47 0.00 0.01 0.09 20.15* 0.02 20.05 20.02 0.08 0.03 20.01 0.0415 Media attention to the CEO 2.97 7.24 20.03 0.08 20.12* 20.16* 20.05 20.07 0.37* 0.65* 0.80* 20.11* 20.0716 CEO celebrity 0.00 0.05 0.02 20.01 20.02 20.03 20.01 20.02 0.10* 0.10* 0.03 0.04 20.0317 CEO background (sales & marketing) 0.40 0.49 20.02 0.00 0.06 0.06 0.13* 0.13* 20.18* 0.03 20.05 20.39* 20.0818 CEO background (finance) 0.14 0.35 0.04 0.15* 20.04 0.06 20.05 20.06 0.09* 0.01 0.05 0.19* 20.0519 Number of statements WSJ 7.70 8.78 20.14* 0.08 20.06 20.12* 20.12* 20.11* 0.29* 0.32* 0.26* 0.00 20.0520 Number of statements NYT 4.55 8.43 20.07 0.08 20.02 20.07 20.10* 20.03 0.21* 0.24* 0.30* 0.08 0.0621 Share of CEO’s words 0.38 0.15 20.05 0.16* 0.01 20.01 0.13* 0.07 20.15* 20.07 20.13* 20.22* 20.21*22 CFO involvement 1.15 2.07 0.02 20.15* 0.09 0.04 20.04 0.05 0.03 0.01 0.02 0.04 0.21*23 CEO future orientation 0.14 0.07 0.05 20.10* 0.13* 0.18* 0.01 0.07 20.15* 20.14* 20.10* 0.09 0.0624 CEO image-based language 0.05 0.01 20.09 0.15* 20.03 20.13* 20.02 0.07 0.07 0.10* 0.06 20.20* 0.0125 CEO comprehensibility 14.32 1.81 20.03 20.12* 0.05 0.10* 20.11* 20.04 20.04 20.13* 20.18* 0.09 20.14*26 CEO numerical language 0.02 0.01 0.02 20.07 0.12* 0.07 0.00 0.01 20.14* 20.17* 20.12* 0.05 0.0327 CEO noncontentual words 0.13 0.13 0.04 20.12* 0.13* 0.11* 0.00 20.03 20.03 20.04 0.05 0.18* 20.0228 CEO optimism 0.31 0.22 20.03 0.12* 0.00 0.02 20.14* 0.07 0.05 0.19* 0.11* 20.04 20.05

23 In our main analysis, we used the I/B/E/S detailed da-tabase on individual analysts’ recommendations instead ofthe I/B/E/S summary database on consensus recommenda-tions,which has been employed in prior studies (e.g., Benner& Ranganathan, 2012). We did so for two reasons. First, thesummary database did not allow us to accurately distinguishrecommendations issued before the call from those issuedafter the call. Second, using the summary database wouldhave caused serious distortion in our models because theclosest consensus estimates before and after the call are notnecessarily calculated from the same group of analysts. Infact, analysts do not issue new recommendations as oftenas they issue EPS forecasts. As such, analysts are frequentlyincluded in consensus estimates before the call but not inconsensus estimates after the call, and vice versa.

24 An analysis that used the median change in recom-mendations before and after the call to measure analysts’favorability, which is comparable to our measure based onEPS forecasts, supports our findings.


included the debt-to-equity ratio, dividends per share,liquidity,25 and cash flow from operating activities inthe quarter preceding the focal conference call (in bil-lion USD) as well as the cumulated standardized ab-normal returns in the 10 days before the call.26We alsoincluded the number of shares traded on the day of thecall relative to the number of shares outstanding(Jegadeesh, Kim, Krische, & Lee, 2004).

CEO controls. We included all CEO controlsfound in Analysis I.

Analyst controls. We accounted for herding ef-fects by including the number of analysts followingeach firm, defined as the number of analysts issuingat least one EPS forecast for the firm in the corre-sponding year (Fanelli et al., 2009).

Conference-call controls. We used all conference-call controls included in Analysis I.

Econometric Approach in Analysis II

Similar to Analysis I, we used a robust firm fixedeffects estimator with the Huber–White standard

error correction. We again mean-centered the com-ponents of the interaction term.

RESULTS

Descriptive Statistics

Table 2 presents means, standard deviations, andcorrelations among the variables for Analysis I. Thestrong correlations betweenmedia attention to the firmand the number of press releases (0.87), and betweenmedia attention to the firm and media attention to theCEO (0.80) can be attributed to firm size (Pollock &Rindova, 2003). This is further corroborated by thecorrelations between firm size and the number of pressreleases (0.63), and between firm size and media at-tention to the firm (0.57).

Table 3 presents means, standard deviations, andcorrelations for Analysis II.27

TABLE 2(Continued)

12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

1.0020.17* 1.000.00 20.70* 1.000.07 20.18* 0.03 1.00

20.05 0.05 20.03 0.01 1.0020.15* 0.05 0.07 20.05 20.04 1.0020.01 0.02 20.10* 0.02 0.12* 20.11* 1.0020.05 20.05 0.06 0.23* 20.01 0.00 20.03 1.000.06 20.04 0.04 0.26* 20.01 20.10* 0.08 0.17* 1.00

20.16* 0.03 0.12* 20.10* 0.06 0.36* 0.05 20.12* 20.08 1.000.15* 0.01 20.06 0.00 20.01 20.05 20.04 0.05 0.03 20.44* 1.000.12* 0.06 20.05 20.04 0.04 20.13* 0.05 0.01 0.00 20.17* 0.13* 1.00

20.18* 20.03 0.04 0.07 20.09 0.10* 20.07 0.02 20.06 0.09 20.05 20.37* 1.000.02 0.08 20.09 20.14* 0.02 0.00 20.04 20.04 20.14* 0.01 20.20* 0.16* 20.23* 1.000.08 0.13* 20.09 20.06 20.06 20.07 0.01 20.09 0.08 20.05 20.02 0.26* 20.15* 20.04 1.000.08 0.01 0.04 0.08 0.00 20.11* 0.05 0.03 0.08 20.37* 0.15* 0.39* 20.45* 0.16* 0.40* 1.000.02 20.09 0.10* 0.07 0.01 0.08 20.06 0.08 0.03 20.07 20.25* 20.15* 0.12* 0.21* 20.07 0.03 1.00

Notes: Dummies for running quarters are not included in this table. n 5 449.*p, .05

25 We measured liquidity as the ratio of cash and short-term investments to assets.

26 The results were robust to changes in the timeframesof cumulative returns.

27 Table 3 shows the descriptive statistics for the samplethat uses the EPS-based measure of analysts’ favorability.The corresponding table for the recommendation-basedmeasure, which can be requested from the authors, showshighly comparable means, standard deviations, and cor-relations. We also conducted the same checks for multi-collinearity in that analysis and derived similar results.


To test forpotentialmulticollinearity,wecalculatedvariance inflation factors (VIFs). For Analysis I, mediaattention to the firm, number of press releases, andmedia attention to the CEO had VIFs greater than 3(8.35,6.60, and3.04, respectively),while themeanVIFamounted to 2.03. Although all VIFs were well below10, we reran our models after dropping these threevariables and found that the results were not materi-ally affected. For Analysis II, media attention to thefirm, firm size, number of press releases, CEO con-tender, andCEOoutsiderhadVIFshigher than3 (8.36,6.71, 6.03, 3.62, and 3.37, respectively; mean VIF 52.2). We reran all models after dropping these vari-ables, but our results were unaffected.

Regression Models

In Tables 4 and 5, we present three models foreach of the three dependent variables: one with thecontrol variables, one that includes CEOs’ use ofmetaphorical communication, and one that includesthe interaction of CEOs’ use of metaphorical com-munication with the negative earnings surprise andpositive earnings surprise splines. As the results areconsistent across models, we only interpret the fullmodels.Model 3 in Table 4 presents the results of thetest of the impact of CEOs’ use of metaphoricalcommunication on journalists’ favorability.28 In thatmodel, the coefficient of CEOs’ use of metaphoricalcommunication is positive and significant (p,0.01).This provides support for Hypothesis 1, which sug-gests that increases in a CEO’s metaphorical com-munication render journalists more benevolenttoward the CEO’s firm.

Models 3 and 6 in Table 5 present the results of thefixed effects panel models designed to test the effectof CEOs’ use of metaphorical communication on an-alysts’ favorability. Model 3 includes the measurebased on EPS forecasts, while Model 6 includes themeasure based on I/B/E/S recommendation data. Inboth models, the coefficient of CEOs’ use of meta-phorical communication is negative and significant(p , 0.05). This provides support for Hypothesis 2,which predicts that firms receive less favorable ana-lyst assessments if CEOs use more metaphoricalcommunication in a given conference call.

Finally, we tested Hypotheses 3a and 3b, whichpredict that the effects of CEOs’ use of metaphorical

communication on infomediaries’ appraisals are am-plified by increasing degrees of negative earningssurprises. We find support for both hypotheses. Asindicated in Model 3 in Table 4 and in Models 3 and6 in Table 5, negative earnings surprises amplifythe association between metaphorical communicationand infomediaries’ (un-)favorability. This is not thecase for positive earnings surprises. Figure 1a visual-izes the moderating effect of negative earnings sur-prises (plus or minus 0.25 SD from the mean) on therelation between CEOs’ use of metaphorical commu-nication (mean-centered) and the favorability of jour-nalists’ reporting. Clearly, the marginal effect of CEOs’useofmetaphorical communication increases themorefirm earnings negatively deviate from expectations.Figure1b shows this interactive effect for theEPS-basedmeasure of analysts’ assessments and suggests a greatermarginal negative effect of metaphorical communica-tion in the case of increasing earnings disappointment.For instance, if the level of metaphorical communica-tion used by the CEO is at 0.01, the favorability of ana-lysts’ evaluations is approximately 0.37 units lower inthe case of a major earnings disappointment (–0.16)than in a situation with a minor earnings disappoint-ment (0.21).29Figure1cshowsacorrespondingeffect forthe recommendation-based measure of analysts’ favor-ability. Notably, we conducted a simple slope analysisfor all interactions (Aiken &West, 1991; Cohen, Cohen,West, &Aiken, 2003). For both journalists and analysts,the test indicated significant effects for low and highlevels of CEO’s metaphorical communication in situa-tions involving a negative earnings surprise. This fur-ther supports our findings.

In order to further ensure the validity of our results,we conducted an extensive set of robustness checks.These involved, for example, the exclusion and in-clusion of covariates, the reevaluationofAnalysis II atthe level of the individual security analyst, and theimplementation of a set of preliminary endogeneitytests. See Appendix A for additional details.

DISCUSSION

Our study shows that CEOs’ use of metaphoricalcommunication shapes how their firms are viewed byinfomediaries. In addition, our results indicate that this

28 As heteroskedasticity suggests that the error terms arenot normally distributed and, thereby, violate a core as-sumption of the F-test, we follow Stata and do not report Fstatistics. Instead, we conducted a log-likelihood ratio test.

29 Assume a share price of USD 100 and an EPS forecastof USD 3 before the call. This number predicts that, ata level of metaphorical communication of 0.01, the EPSforecast would drop after the call by approximately 5.3%,to USD 2.84, in the case of a major earnings disappoint-ment, ceteris paribus.


effect is not uniform, but depends on the idiosyncraticsocial situations of different types of infomediaries.While journalists reportmore favorably about firms themore their CEOs rely onmetaphorical communication,analysts issue more critical forecasts the more CEOsuse this typeof communication.Moreover, theeffectsof CEOs’ use of metaphorical communication onjournalists’ and analysts’ favorability grow the morefirm performance fails to meet expectations.

Classical Rhetoric and CEOs’ Communicationwith Infomediaries

Our research contributes to the growing debatesurrounding the power of CEOs’ words to influenceinfomediaries (e.g., Fanelli et al., 2009;Westphal et al.,2012) by directing attention to the implications ofsome of the most widely taught elements of classical

rhetoric. Most prior work on executive communica-tion (Westphal et al., 2012) and impression manage-ment (e.g., Elsbach, 1994) has focused on the use ofspecific types of words, such as image-based words(Emrich, Brower, Feldman, & Garland, 2001) andemotional words (Guo, 2014); on the content of com-munication, such as prosocial claims (McDonnell &King, 2013); or on a combination of the two (Cartonet al., 2014). Although these studies have revealedimportant insights, few researchers have investi-gated the vital notion that CEOs’ sensegiving towardinfomediaries is not only about choosing certaintypes of words and presenting certain types of firm-related information, but also about using rhetoricalinstruments to provide frames that direct audiences’attention and interpretations (Amernic et al., 2007;Clatworthy & Jones, 2003). We address this gap bybuilding ona largebodyof studies on cognition (Gioia,

TABLE 3Correlations and Descriptive Statistics of Analysis II

Variable Mean SD 1 2 3 4 5 6 7

1 Favorability of analysts’ evaluations 0.20 1.26 1.002 CEO’s use of metaphorical

communication0.01 0.01 20.07* 1.00

3 Prior firm performance change 0.02 0.12 0.12* 20.01 1.004 Prior firm performance volatility 0.11 0.13 0.02 20.08* 0.12* 1.005 Negative earnings surprise 0.07 0.20 20.32* 20.08* 0.00 0.09* 1.006 Positive earnings surprise 0.19 0.34 0.24* 20.07 0.13* 0.18* 20.20* 1.007 Firm size 7.85 1.88 20.03 0.22* 20.14* 20.32* 20.27* 20.20* 1.008 Number of press releases 290.63 466.26 0.00 0.10* 20.17* 20.21* 20.17* 20.10* 0.65*9 Media attention to the firm 91.33 194.87 0.00 0.15* 20.14* 20.18* 20.13* 20.09* 0.57*

10 Debt-to-equity ratio 1.80 2.55 0.01 0.01 0.15* 0.05 0.01 0.04 20.14*11 Dividends per share 0.05 0.13 20.04 0.13* 0.02 20.07* 20.11* 20.15* 0.45*12 Liquidity 0.37 0.24 0.04 20.20* 0.22* 0.40* 0.21* 0.12* 20.64*13 Cash flow from operating activities 0.89 2.42 0.00 0.05 20.09* 20.14* 20.12* 20.13* 0.59*14 Abnormal returns 20.05 3.33 0.09* 0.00 20.02 20.07 20.07 20.02 0.0415 Number of shares traded 23.46 28.28 20.01 20.06 0.12* 0.09* 0.27* 0.05 20.17*16 CEO age 53.03 6.06 0.03 0.01 20.02 20.07 20.04 20.12* 0.18*17 CEO tenure 3.74 3.62 0.01 20.09* 0.02 20.02 20.01 0.04 20.13*18 CEO duality 0.44 0.50 20.04 20.01 20.01 20.05 20.12* 20.03 0.26*19 Contender 0.48 0.50 20.01 20.03 0.04 0.13* 0.01 0.05 20.0720 Outsider 0.38 0.49 0.02 0.02 0.03 20.09* 0.03 20.07 20.0621 Media attention to the CEO 1.79 5.90 20.01 0.13* 20.09* 20.13* 20.06 20.07 0.36*22 CEO celebrity 0.00 0.04 0.00 20.01 20.02 20.02 20.01 20.02 0.09*23 CEO background (sales &

marketing)0.37 0.48 0.00 0.02 0.01 20.02 0.04 0.06 20.06

24 CEO background (finance) 0.16 0.37 0.03 0.12* 20.03 0.05 0.01 20.02 0.0225 Number of analysts following 19.19 10.08 0.03 0.12* 20.09* 20.19* 20.21* 20.12* 0.61*26 Share of CEO’s words 0.42 0.16 20.01 0.04 0.00 0.04 0.09* 0.06 20.29*27 CFO involvement 0.96 1.12 20.03 20.13* 0.04 20.04 20.03 0.03 0.13*28 CEO future orientation 0.15 0.07 0.04 20.16* 0.12* 0.12* 0.08* 0.08* 20.19*29 CEO image-based language 0.05 0.01 0.00 0.22* 20.06 20.19* 20.10* 0.00 0.11*30 CEO comprehensibility 14.32 1.82 0.10* 20.06 0.02 0.07 20.08* 20.04 0.0231 CEO numerical language 0.02 0.01 0.04 20.10* 0.10* 20.01 20.01 0.00 20.16*32 CEO noncontentual words 0.13 0.12 0.04 20.14* 0.13* 0.14* 0.06 0.00 20.0433 CEO optimism 0.30 0.21 0.06 0.09* 0.00 20.04 20.09* 0.02 0.05


1986) and linguistics (Steen, 2011) that have high-lighted classical rhetorical devices, especially meta-phorical communication, as ways of persuadingcritical audiences by reducing complexity and for-mulating “sticky”messages (Heath & Heath, 2007). Inso doing, we draw attention to the potentially crucialrole of the larger canon of classical rhetoric (Corbett &Connors, 1998) for our understanding of CEOs andtheir quest to manage the legitimacy of their firms.

Our study is also unique because it develops noveltheoretical explanations for why CEOs’ rhetoric affectsinfomediaries’ appraisals. By integrating general theoryon social cognition (Fiske & Taylor, 2017) with specifictheory on infomediaries’ socially situated sensemaking(Wiesenfeld et al., 2008), we emphasize that infome-diaries interpret firms’ and CEOs’ public language inhighly charged and influential social contexts. This

perspective is particularly useful because it allows usto view CEOs’ communication through the idiosyn-cratic lenses of the various types of infomediaries—individuals who are similar in that they all workunder tight constraints and all aim to serve theiraudiences, but who differ because they pass senseon to different recipients and because they are so-cially situated in fundamentally different thoughtworlds (Lamin & Zaheer, 2012).

Metaphorical Communication as a Metaphor:Understanding CEOs’ Rhetorical Dilemmas

Ourstudyalsoprovidesasystematic, contextualizedpicture of CEOs’ use of metaphorical communication.While research in various domains has increasinglysuggested that communicators need to consider the

TABLE 3(Continued)

Variable 8 9 10 11 12 13 14 15 16 17 18

1 Favorability of analysts’ evaluations2 CEO’s use of metaphorical

communication3 Prior firm performance change4 Prior firm performance volatility5 Negative earnings surprise6 Positive earnings surprise7 Firm size8 Number of press releases 1.009 Media attention to the firm 0.87* 1.00

10 Debt-to-equity ratio 0.01 0.04 1.0011 Dividends per share 0.21* 0.20* 20.17* 1.0012 Liquidity 20.33* 20.29* 0.14* 20.25* 1.0013 Cash flow from operating activities 0.59* 0.57* 20.09* 0.36* 20.28* 1.0014 Abnormal returns 0.03 20.01 0.02 20.04 20.01 0.03 1.0015 Number of shares traded 20.13* 20.10* 0.21* 20.13* 0.15* 20.14* 0.01 1.0016 CEO age 0.01 0.05 20.07* 0.33* 20.10* 0.15* 0.01 20.08* 1.0017 CEO tenure 20.10* 20.12* 0.03 20.05 0.14* 20.09* 0.01 0.15* 0.17* 1.0018 CEO duality 0.17* 0.15* 20.02 0.16* 20.17* 0.21* 20.03 0.07 0.17* 0.27* 1.0019 Contender 20.14* 20.16* 20.03 0.04 0.13* 20.04 20.09* 0.10* 0.07 20.29* 20.08*20 Outsider 0.07* 0.04 0.06 20.10* 20.11* 20.02 0.08* 20.07 20.10* 0.02 20.0421 Media attention to the CEO 0.63* 0.79* 0.03 0.10* 20.18* 0.39* 20.02 20.05 20.05 20.10* 0.0622 CEO celebrity 0.12* 0.04 20.02 0.04 20.04 0.18* 0.02 20.01 0.03 20.03 20.0423 CEO background (sales &

marketing)0.09* 0.01 0.00 20.17* 0.12* 20.05 0.03 20.01 20.45* 20.14* 20.09*

24 CEO background (finance) 0.01 0.07 20.06 0.19* 0.02 0.11* 0.04 0.00 0.19* 20.09* 20.0525 Number of analysts following 0.53* 0.52* 0.10* 0.07 20.21* 0.34* 0.07* 0.00 20.02 20.01 0.0526 Share of CEO’s words 20.10* 20.13* 20.16* 20.06 0.10* 20.06 0.06 0.00 20.24* 20.20* 20.19*27 CFO involvement 0.00 0.01 0.03 0.08* 20.10* 20.01 20.06 20.03 0.14* 0.18* 0.17*28 CEO future orientation 20.14* 20.10* 20.07 20.02 0.18* 20.11* 20.04 0.08* 0.07 0.15* 0.10*29 CEO image-based language 0.06 0.02 20.01 20.05 20.15* 20.04 0.00 20.05 20.12* 20.03 20.15*30 CEO comprehensibility 20.05 20.11* 20.03 20.04 0.09* 20.08* 20.04 0.00 0.05 20.14* 0.0131 CEO numerical language 20.14* 20.09* 0.03 20.09* 20.13* 20.12* 20.02 0.13* 0.02 0.04 0.0432 CEO noncontentual words 20.02 0.07 0.05 0.02 0.03 20.03 20.02 0.19* 0.14* 0.10* 0.0733 CEO optimism 0.14* 0.05 0.08* 20.13* 0.00 0.02 0.02 20.06 0.00 20.02 20.03


diverging needs of their intended audiences whenusing metaphorical communication (Black, 1962;Dunbar, 1995; Liu, 2002), few studies in the manage-ment domain have proposed that the effects of meta-phorical communication could bemore complex thantypically portrayed in the practitioner-oriented litera-ture (Cornelissen et al., 2011; Konig et al., 2017;Ramsay,2004).To thebest of ourknowledge,our studyis the first to build theory and use systematic evidenceto explore how and why the use of metaphoricalcommunication might be particularly intricate in thecontext of CEOs’ strategic public language.

Part of whatmakes our findings insightful is that theyreveal the rhetoricaldilemma thatCEOsfacewhenusing

metaphorical communication. This dilemma is rootedin the fact that CEOs often target their communicationat different infomediary groups simultaneously—in ourcase, journalists and analysts. Given that it is institu-tionally difficult to separate these groups, CEOs have tochoosebetween twosuboptimal alternatives: either theyjeopardize analysts’ benevolence by using more meta-phoricalcommunicationor theyforgo theopportunity togarner more positive journalist reporting about the firmby using less metaphorical communication. Thus, ona broader scale, metaphorical communication mightserve as a metaphor for the larger phenomenon of therhetorical dilemmas in executive communication thatemerge from the fact that executive communication

TABLE 3(Continued)

19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

1.0020.76* 1.0020.12* 0.04 1.000.04 20.03 0.02 1.000.05 0.03 0.01 20.03 1.00

0.00 20.06 0.03 0.09* 20.17* 1.0020.07 20.08* 0.37* 0.03 0.12* 0.01 1.0020.02 0.17* 20.09* 0.04 0.27* 0.07 20.31* 1.0020.02 20.07 20.01 20.02 20.13* 20.05 0.11* 20.58* 1.000.06 20.08* 20.04 0.02 20.07 0.03 20.13* 20.08* 0.10* 1.00

20.05 0.06 0.03 20.07 0.11* 20.05 0.12* 0.08* 20.08* 20.38* 1.000.19* 20.18* 20.08* 0.01 0.01 20.08* 0.06 20.03 20.07* 0.10* 20.16* 1.000.11* 20.07* 20.04 20.05 20.07 0.01 20.23* 20.01 0.07 0.23* 20.11* 20.11* 1.000.04 20.02 0.09* 20.01 20.16* 0.05 20.06 20.31* 0.20* 0.41* 20.45* 0.07 0.38* 1.000.01 0.01 0.03 0.01 0.12* 20.13* 0.11* 20.12* 20.03 20.08* 0.09* 0.20* 20.03 0.03 1.00

Notes: Dummies for running quarters are not included in this table. n 5 624.*p, .05


TABLE 4Results of Robust Fixed Effects Analysis of CEOs’ Use of Metaphorical Communication on the

Favorability of Journalists’ Reporting

Journalists’ Favorability

Variable (1) (2) (3)

Prior firm performance change 20.261 20.311 20.271

(0.15) (0.15) (0.16)Prior firm performance volatility 20.10 20.06 20.04

(0.27) (0.26) (0.27)Negative earnings surprise† 0.03 0.05 0.21

(0.15) (0.15) (0.15)Positive earnings surprise† 0.07 0.06 0.08

(0.08) (0.08) (0.08)Firm size 20.04 20.03 20.03

(0.05) (0.05) (0.05)Number of press releases 0.00 0.00 0.00

(0.00) (0.00) (0.00)Media attention to the firm 20.00 20.00 20.00

(0.00) (0.00) (0.00)CEO age 20.00 20.00 20.00

(0.00) (0.00) (0.00)CEO tenure 20.00 20.00 20.00

(0.01) (0.01) (0.01)CEO duality 20.04 20.04 20.05

(0.04) (0.04) (0.04)Contender 0.01 0.02 0.02

(0.04) (0.04) (0.04)Outsider 0.00 0.01 20.01

(0.06) (0.05) (0.06)Media attention to the CEO 0.00 0.00 0.00

(0.00) (0.00) (0.00)CEO celebrity 0.20 0.19 0.18

(0.12) (0.12) (0.12)CEO background (sales & marketing) 20.04 20.04 20.04

(0.05) (0.05) (0.05)CEO background (finance) 20.01 20.02 20.01

(0.05) (0.05) (0.05)Number of statements WSJ 20.01*** 20.01*** 20.01***

(0.00) (0.00) (0.00)Number of statements NYT 20.00 20.00 20.00

(0.00) (0.00) (0.00)Share of CEO’s words 20.31 20.37 20.391

(0.22) (0.23) (0.22)CFO involvement 0.00 0.01 0.01

(0.01) (0.01) (0.01)CEO future orientation 0.17 0.17 0.18

(0.25) (0.26) (0.26)CEO image-based language 22.781 23.01* 22.85*

(1.45) (1.42) (1.39)CEO comprehensibility 20.02 20.02 20.02

(0.01) (0.01) (0.01)CEO numerical language 3.40 3.41 3.06

(2.74) (2.57) (2.50)CEO noncontentual words 20.28 20.26 20.28

(0.19) (0.19) (0.18)CEO optimism 0.01 20.02 20.01

(0.10) (0.10) (0.09)CEO’s use of metaphorical communication† 6.16* 7.64**

(2.82) (2.75)


is almost always received by multiple audiences withpotentially contradictory interests.

Connecting Disconnected Debates in Researchon Infomediaries

Our study also has more general implications forresearch on firms’ relationswith infomediaries.Mostimportantly, it is the first to compare the effects ofone facet of CEO communication on two groups ofinfomediaries. In this respect, our research extendsand challenges prior studies that have focused onsingle audiences (e.g., Fanelli et al., 2009; Zavyalovaet al., 2012). In particular, it provides novel expla-nations for why findings on firm-intermediary dis-course cannot necessarily be generalized acrossaudiences, thereby adding to research that highlightsthe complex, potentially paradoxical effects of firms’communication with diverse constituents (e.g., Lamin& Zaheer, 2012; Zavyalova et al., 2016).

Finally, by showing that CEOs’ rhetoric influencesthe benevolence of infomediaries, we further openup the black box of behavioral tendencies in info-mediaries’ evaluations (Mokoaleli-Mokoteli, Taffler,& Agarwal, 2009). For instance, while the extant re-search has explained why analysts are often overlyoptimistic when issuing evaluations (Sedor, 2002),scholars are still unclear as to why analysts aresometimes overly pessimistic in their assessments(Doukas, Kim, & Pantzalis, 2002). Our findings—including those on the interactive effects of meta-phorical communication and negative earnings

surprises—indicate that the influence of rhetoricaldevices for constituent-minded sensemaking arerelevant for explaining such behavior.

Practical Implications

There are important practical implications ofour research. Perhaps most importantly, we advisecorporate leaders, along with their coaches andspeechwriters, to view metaphorical communicationand other elements of classical rhetoric as vital butequivocal levers for influencing infomediaries’ eval-uations. To manage the trade-offs involved in usingmetaphorical communication when interacting withinfomediaries, CEOs might need to consider whichgroup of infomediaries is most important at a giventime and tailor their rhetoric to the preferences of thataudience. Such considerations are especially impor-tant in times of poor firm performance.

FUTURE RESEARCH

We acknowledge the limitations of this study,which, in turn, point to promising avenues for futureresearch. Most notably, given the interpretativecomplexities inherent inouranalysis,wecouldonlyusepreliminarymeasures to control for the degree towhichCEOs’ metaphorical communication was aligned withthe message that was to be conveyed (Aristotle, n.d.;Booth,1978).This is importantbecause,according to theinteraction view of metaphor (Black, 1962), audiencesmake sense of metaphorical language by drawing asso-ciations between the structures of the source and target

TABLE 4(Continued)

Journalists’ Favorability


CEO’s use of metaphorical communication 53.57*3 Negative earnings surprise (22.49)

CEO’s use of metaphorical communication 20.163 Positive earnings surprise (11.23)

n 449 449 449R2 within 0.15 0.16 0.18Log likelihood 275.99 269.16 263.74LR (likelihood ratio) x2 against null model 70.23*** 77.99*** 88.83***

Notes: Robust standard errors in parentheses. Models include dummies for eachquarter from 2002 to 2011.

† Variable is centered at its mean.1p , .1*p , .05

**p , .01***p , .001


TABLE 5Results of Fixed–Effects Analysis of CEOs’Use of Metaphorical Communication on the Favorability of Analysts’ Evaluations

Analysts’ Favorability

EPS forecast measurea Recommendation measureb

Variable (1) (2) (3) (4) (5) (6)

Prior firm performance change 0.40 0.50 0.49 20.50 20.42 20.25(0.53) (0.52) (0.52) (0.49) (0.53) (0.47)

Prior firm performance volatility 20.621 20.671 20.60 20.50 20.53 20.59(0.35) (0.36) (0.36) (0.65) (0.65) (0.63)

Negative earnings surprise† 22.30*** 22.29*** 22.60*** 20.34 20.23 20.30(0.38) (0.37) (0.31) (0.42) (0.40) (0.40)

Positive earnings surprise† 0.67* 0.68* 0.70** 0.29* 0.27* 0.35*(0.27) (0.27) (0.26) (0.13) (0.13) (0.16)

Firm size 20.431 20.411 20.401 0.02 0.12 0.11(0.23) (0.22) (0.22) (0.28) (0.26) (0.26)

Number of press releases 0.00 0.00 0.00 20.00 20.00 20.00(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

Media attention to the firm 0.00 0.00 0.00 20.00 20.00 20.00(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)

Debt-to-equity ratio 0.02 0.02 0.02 0.01 0.02 0.03(0.01) (0.01) (0.01) (0.04) (0.04) (0.03)

Dividends per share 20.761 20.711 20.791 20.40 20.04 20.01(0.39) (0.41) (0.41) (1.23) (1.08) (1.17)

Liquidity 20.781 20.791 20.77 0.961 1.05* 1.061

(0.44) (0.43) (0.47) (0.56) (0.52) (0.56)Cash flow from operating activities 20.02 20.02 20.02 0.00 20.00 0.00

(0.02) (0.02) (0.02) (0.00) (0.00) (0.00)Abnormal returns 0.03 0.03 0.03 0.02 0.02 0.02

(0.02) (0.02) (0.02) (0.02) (0.02) (0.02)Number of shares traded 0.01 0.01 0.00 0.00 0.00 0.00

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)CEO age 20.01 20.00 20.01 20.01 20.00 0.00

(0.01) (0.01) (0.01) (0.01) (0.02) (0.02)CEO tenure 20.02 20.02 20.02 0.051 0.041 0.04

(0.02) (0.02) (0.02) (0.02) (0.02) (0.02)CEO duality 20.15 20.14 20.11 0.19 0.27 0.29

(0.15) (0.15) (0.15) (0.22) (0.19) (0.19)Contender 20.08 20.10 20.08 0.22 0.20 0.25

(0.21) (0.21) (0.21) (0.32) (0.31) (0.30)Outsider 0.01 0.03 0.08 20.02 20.15 20.14

(0.21) (0.22) (0.21) (0.27) (0.27) (0.29)Media attention to the CEO 20.01 20.01 20.01 0.00 0.01 0.01

(0.01) (0.01) (0.01) (0.01) (0.01) (0.01)CEO celebrity 20.72 20.73 20.64 0.04 0.05 0.00

(0.59) (0.60) (0.60) (0.64) (0.59) (0.60)CEO background (Sales & Marketing) 20.07 20.05 20.05 0.391 0.31 0.27

(0.15) (0.15) (0.16) (0.23) (0.23) (0.24)CEO background (Finance) 0.21 0.22 0.19 20.09 20.08 20.06

(0.17) (0.18) (0.18) (0.26) (0.25) (0.28)Number of analysts following 20.01 20.01 20.01 0.01 0.01 0.01

(0.01) (0.01) (0.01) (0.02) (0.01) (0.02)Share of CEO’s words 20.29 20.27 20.39 20.72 20.50 20.65

(0.55) (0.55) (0.54) (0.61) (0.61) (0.59)CFO involvement 20.04 20.05 20.05 20.02 20.02 20.02

(0.04) (0.04) (0.04) (0.02) (0.02) (0.02)CEO future orientation 20.09 20.09 20.24 21.00 21.17 21.38

(1.13) (1.12) (1.08) (1.28) (1.24) (1.21)


domains (Zashin & Chapman, 1974). However, somemetaphors might be perceived as so “bizarre” that au-dienceswill struggle tomake such associations.Whilewe did not come across any such examples,30 we seeampleopportunities for future research into the impactof specific characteristics of CEOs’metaphorical com-munication in, for instance, highly diverse culturalcontexts (Konig et al., 2017; Liu, 2002). As part of theseendeavors, scholars might also find ways to build onand improve our research design. In particular, we

envision research that automatizes the detection andclassification of CEOs’ metaphorical communicationby, for example, developing dictionaries for specificsource domains, such as sports, journeys, or violence(Lakoff & Johnson, 1980). The literature on appliedlinguistics (e.g., Cameron & Maslen, 2010) might pro-vide useful guidance in this regard.

Furthermore, scholars might draw a more com-prehensive picture of CEO communication withinfomediaries and the role of metaphorical commu-nication in that context by analyzing other discur-sive vehicles, such as interviews, public speeches,corporate presentations, or roadshows (Whittington,Yakis-Douglas, & Ahn, 2016). Incorporating suchsources would enhance our understanding of howinfomediaries perceive CEO communication andthe underlying mechanisms. Relatedly, there is am-ple scope for research into how CEOs deliver com-munication by including such aspects as facial

TABLE 5(Continued)

Analysts’ Favorability

EPS forecast measurea Recommendation measureb

Variable (1) (2) (3) (4) (5) (6)

CEO image-based language 21.09 20.05 0.76 1.46 2.85 3.59(3.64) (3.87) (3.81) (8.70) (8.53) (8.39)

CEO comprehensibility 0.09** 0.09** 0.08** 0.04 0.03 0.03(0.03) (0.03) (0.03) (0.04) (0.04) (0.04)

CEO numerical language 13.381 13.771 14.63* 23.56 26.71 29.71(7.09) (6.98) (7.18) (11.04) (10.64) (10.78)

CEO noncontentual words 0.28 0.22 0.29 20.96 21.03 21.04(0.66) (0.64) (0.63) (0.81) (0.76) (0.73)

CEO optimism 0.39 0.45 0.35 20.03 0.04 20.06(0.34) (0.34) (0.34) (0.29) (0.28) (0.27)

CEO’s use of metaphorical communication† 213.76* 215.42* 215.791 219.35*(6.41) (5.88) (8.55) (7.79)

CEO’s use of metaphorical communication 2120.25** 2107.30*3 Negative earnings surprise (35.51) (50.80)

CEO’s use of metaphorical communication 14.96 34.041

3 Positive earnings surprise (24.32) (19.40)

n 624 624 624 270 270 270R2 within 0.29 0.29 0.30 0.29 0.31 0.34Log likelihood 2907.09 2905.13 2899.91 2223.56 2220.66 2214.84LR x2 against null model 209.86*** 213.79*** 224.22*** 92.98*** 98.79*** 110.43***

Notes: Robust standard errors in parentheses. Models include dummies for each quarter from 2002 to 2011.a Based on 6,969 before or after the conference call comparisons of individual analysts’ EPS forecasts.b Based on 393 before or after the conference call comparisons of individual analysts’ recommendations.† Variable is centered at its mean.1p , .1*p , .05

**p , .01***p , .001

30 We discussed several metaphors as potentially bi-zarre, but then decided they were sufficiently compre-hensible. Such metaphors included, for instance: “Theycan sort of try and buy as opposed to choke down a hugehairball” (Carol Bartz, ACAD/Q3/2006), and “I don’t be-lieve you’re going to see very many people with crystalballs that don’t have lots of cloud and cotton and fuzz inthem” (Jack London, CACI/Q3/2007).


expressions, gestures, and tone of voice (Cornelissen,Clarke, & Cienki, 2012; Den Hartog & Verburg, 1997;Wenzel & Koch, 2017).

We also see opportunities to examine the effects ofCEOs’ use of other rhetorical devices on infomediaries.We chose to focus on metaphorical communicationbecause linguists suggest that metaphors are particu-larly suitable for framing complexmessages under timeconstraints and ambiguous conditions (e.g., Sopory &Dillard, 2002), aspects that are at the heart of CEOs’communication with infomediaries. But could info-mediaries’ social cognitions also affect the outcomesassociatedwithotherpartsofCEOs’ rhetoric?Moreover,given infomediaries’ rhetorical biases, could the ef-fect of metaphorical communication be amplified byother aspects of CEOs’ communication? The signifi-cantly positive effect of CEOs’ use of numerical lan-guage on analysts’ favorability that we observe (seeModel3 inTable5)mightpoint in thisdirection (Henry,2008). Thus, future research should address thesequestions—their answersmight provide cues as to how

firmscanbest approach thedilemmaofcommunicatingwith diverse audiences.

Finally, subsequent research could extend the viewspresented here by studying the potentially divergenteffects of CEOs’ use of metaphorical communicationand other dimensions of classical rhetoric on otheraudiences. Obviously, it would be worthwhile to ex-tend our analysis to other types of infomediaries, suchas customer-advocacy groups or rating agents, whichthemselves have other audiences and, thus, might re-spond differently to metaphorical communication.Such extensions might also more directly illuminatethe mechanisms through which CEOs’ use of meta-phorical communication affects infomediaries’ ap-praisals, for which we only provide preliminaryevidence. Moreover, it would be interesting to studythe effects of metaphorical communication on differ-ent shareholder groups (Hayward & Fitza, 2016;Whittington et al., 2016), such as institutional in-vestors, private investors, family investors, and pro-fessional investors. Each of these groups might have

FIGURE 1The Interaction Effects of CEO’s Use of Metaphorical Communication and Negative Earnings Surprise on the

Favorability of Infomediaries’ Evaluations

Negative earnings surprise (–0.25 SD)*

Negative earnings surprise (+0.25 SD)*

* We used 0.25 SD in order to remain within the range of the “negative earnings surprise” variable

–0.10

0.00

0.10

0.20

0.30

0.40

–0.010 –0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030

CEO’s use of metaphorical communication (mean-centered)

Journalist’s favorability

–0.50

–0.30

–0.10

0.10

0.30

0.50

–0.010 –0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030


Analyst’s favorability (EPS-forecast measure)

–0.70

–0.55

–0.40

–0.25

–0.10

0.05

0.20

0.35

–0.010 –0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030


Analyst’s favorability (recommendation measure)

A

B C


idiosyncratic, socially situated ways of interpretingCEOs’ rhetoric, thereby giving rise to additional vexingparadoxes. We also call for more research on the effectof CEOs’ use of metaphorical communication on au-diences within firms (Konig et al., 2017). Future re-search might, for example, reveal that metaphoricalcommunication polarizes, rather than unifies, organi-zational members and, as such, creates additionalrhetorical dilemmas for corporate leaders.

In conclusion, we hope that our study can serve asa starting point for conversations on leaders’ rhetoricin awide range of research domains and as a first steptoward a more nuanced view of the effects of CEOs’use of metaphorical communication.

REFERENCES

Aiken, L. S., & West, S. G. 1991.Multiple regression: Test-ing and interpreting interactions. Newbury Park, CA:Sage.

Allison, P. D. 1990. Change scores as dependent variablesin regression analysis.SociologicalMethodology, 20:93–114.

Andsager, J. L. 2000. How interest groups attempt to shapepublic opinion with competing news frame. Journal-ism&Mass Communication Quarterly, 77: 577–592.

Antonakis, J., Bendahan, S., Jacquart, P., & Lalive, R. 2010.On making causal claims: A review and recommen-dations. The Leadership Quarterly, 21: 1086–1120.

Antonakis, J., Fenley, M., & Liechti, S. 2011. Can charismabe taught? Tests of two interventions. Academy ofManagement Learning & Education, 10: 374–396.

Antonakis, J., Fenley, M., & Liechti, S. 2012. Learning cha-risma. Transform yourself into the person others wantto follow.Harvard Business Review, 90: 127–130.

Amernic, J., Craig, R., & Tourish, D. 2007. The trans-formational leader as pedagogue, physician, architect,commander, and saint: Five root metaphors in JackWelch’s letters to stockholders of General Electric.Human Relations, 60: 1839–1872.

Aristotle, n.d. Rhetoric. Retrieved from classics.mit.edu/Aristotle/rhetoric.html.

Ball-Rokeach, S. J., & DeFleur, M. L. 1976. A dependencymodel of mass-media effects. Communication Re-search, 3: 3–21.

Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs,K. D. 2001. Bad is stronger than good. Review ofGeneral Psychology, 5: 323–370.

Benford, R. D., & Snow, D. A. 2000. Framing processesand social movements: An overview and assessment.Annual Review of Sociology, 26: 611–639.

Benner, M. J. 2010. Securities analysts and incumbent re-sponse to radical technological change: Evidence fromdigital photography and internet telephony. Organi-zation Science, 21: 42–62.

Benner, M. J., & Ranganathan, R. 2012. Offsetting illegiti-macy: How pressures from securities analysts influenceincumbents in the faceofnewtechnologies.AcademyofManagement Journal, 55: 213–233.

Bergh, D. D., & Fairbank, J. F. 2002. Measuring and testingchange in strategic management research. StrategicManagement Journal, 23: 359–366.

Black, M. 1962. Models and metaphors. Ithaca, NY: Cor-nell University Press.

Block, S. B. 1999. A study of financial analysts: Practiceand theory. Financial Analysts Journal, 55: 86–95.

Booth, W. 1978. Metaphor as rhetoric. Critical Inquiry, 5:49–72.

Brown, L. D. 2001. A temporal analysis of earnings sur-prises: Profits versus losses. Journal of AccountingResearch, 39: 221–241.

Brown, L. D., Call, A. C., Clement, M. B., & Sharp, N. Y.2015. Inside the “black box” of sell-side financial an-alysts. Journal of Accounting Research, 53: 1–47.

Buffett, W. 1985. Chairman’s letter. Berkshire HathawayInc., Annual Report. Omaha: Berkshire Hathaway Inc.Retrieved from http://www.berkshirehathaway.com/letters/1985.html. Accessed May 23, 2016.

Bundy, J., & Pfarrer, M. D. 2014. Reputations in flux: Ex-amining howa firm’smultiple reputations influencereactions to a negative violation. Unpublished Work-ing Paper, Pennsylvania State University and Univer-sity of Georgia.

Burbules,N. C., Schraw,G., &Trathen,W. 1989.Metaphor,idiom, and figuration. Figuration.Metaphor and Sym-bolic Activity, 4: 93–110.

Burns, L. S. 2013. Understanding journalism (2nd ed.).London, U.K.: Sage.

Bushee, B. J., Core, J. E., Guay, W., & Hamm, S. J. W. 2010.The role of the business press as an informationintermediary. Journal of Accounting Research, 48:1–19.

Cameron, L., & Maslen, R. 2010. Metaphor analysis: Re-search practice in applied linguistics, social sci-ences and the humanities. London, U.K.: Equinox.

Carlson, N. 2011. Rolling thunder: AOL CEO Tim Arm-strong’s favorite metaphors and analogies. Retrievedfrom http://www.businessinsider.com/rolling-thunder-aol-ceo-tim-armstrongs-favorite-metaphors-and-analogies-2011-2?IR5T. Accessed May 23, 2016.

Carton, A. M., Murphy, C., & Clark, J. R. 2014. A (blurry)vision of the future: How leader rhetoric about


http://classics.mit.edu/Aristotle/rhetoric.html

http://classics.mit.edu/Aristotle/rhetoric.html

http://www.berkshirehathaway.com/letters/1985.html

http://www.berkshirehathaway.com/letters/1985.html

http://www.businessinsider.com/rolling-thunder-aol-ceo-tim-armstrongs-favorite-metaphors-and-analogies-2011-2?IR&tnqh_x3d;T




ultimate goals influences performance. Academy ofManagement Journal, 57: 1544–1570.

Chatterjee, A., & Hambrick, D. C. 2011. Executive person-ality, capability cues, and risk taking:HownarcissisticCEOs react to their successes and stumbles. Admin-istrative Science Quarterly, 56: 202–237.

Chen, X., & Cheng, Q. 2006. Institutional holdings andanalysts’ stock recommendations. Journal of Ac-counting, Auditing & Finance, 21: 399–440.

Chen, G., Crossland, C., & Luo, S. 2015. Making the samemistake all over again: CEO overconfidence and cor-porate resistance to corrective feedback. StrategicManagement Journal, 36: 1513–1535.

Clatworthy, M., & Jones, M. J. 2003. Financial reporting ofgood news and bad news: Evidence from accountingnarratives. Accounting and Business Research, 33:171–185.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. 2003. Ap-plied multiple regression/correlation analysis forthe behavioral sciences (3rd ed.). Mahwah, NJ:Lawrence Erlbaum.

Corbett, E. P. J., & Connors, R. J. 1998. Classical rhetoricfor the modern student (4th ed.). New York, NY:Oxford University Press.

Cornelissen, J. P., Clarke, J. S., & Cienki, A. 2012. Sense-giving in entrepreneurial contexts: The use of meta-phors in speech and gesture to gain and sustainsupport for novel business ventures. InternationalSmall Business Journal, 30: 213–241.

Cornelissen, J. P., Holt, R., & Zundel, M. 2011. The role ofanalogyandmetaphor intheframingandlegitimizationofstrategic change.Organization Studies, 32: 1701–1716.

Cornelissen, J. P., &Werner,M. D. 2014. Putting framing inperspective: A review of framing and frame analysisacross the management and organizational literature.The Academy of Management Annals, 8: 181–235.

Corts, D. P., & Pollio, H. R. 1999. Spontaneous productionof figurative language and gesture in college lectures.Metaphor and Symbol, 14: 81–100.

Coy, P. 2013. Staggering drunks and fiscal cliffs: HowBloomberg Businessweek uses metaphors in thenews. Retrieved from http://niemanreports.org/articles/staggering-drunks-and-fiscal-cliffs-how-bloomberg-businessweek-uses-metaphors-in-the-news. AccessedMay 23, 2016.

Deephouse, D. L. 2000. Media reputation as a strategic re-source: An integration of mass communication andresource-based theories. Journal ofManagement, 26:1091–1112.

Deephouse, D. L., & Heugens, P. P. 2009. Linking socialissues to organizational impact: The role of infome-diaries and the infomediary process. Journal of Busi-ness Ethics, 86: 541–553.

Den Hartog, D. N., & Verburg, R. M. 1997. Charisma andrhetoric: Communicative techniques of internationalbusiness leaders. The Leadership Quarterly, 8:355–391.

Deuze, M. 2005. What is journalism? Professional identityand ideology of journalists reconsidered. Journalism,6: 442–464.

Doukas, J. A., Kim, C., & Pantzalis, C. 2002. A test ofthe errors-in-expectations explanation of the value/glamour stock returns performance: Evidence fromanalysts’ forecasts. The Journal of Finance, 57:2143–2165.

Dunbar, K. 1995. How scientists really reason: Scientificreasoning in real-world laboratories. In R. J. Sternberg& J. E. Davidson (Eds.), The nature of insight: 365–395.Cambridge, MA: MIT Press.

Elsbach,K.D. 1994.Managing organizational legitimacy inthe California cattle industry: The construction andeffectiveness of verbal accounts. Administrative Sci-ence Quarterly, 39: 57–88.

Emrich, C. G., Brower, H. H., Feldman, J. M., & Garland, H.2001. Images in words: Presidential rhetoric, cha-risma, and greatness. Administrative Science Quar-terly, 46: 527–557.

Entman, R. M. 1993. Framing: Toward clarification ofa fractured paradigm. Journal of Communication, 43:51–58.

Fainsilber, L., & Ortony, A. 1987. Metaphorical uses oflanguage in the expression of emotions. Metaphorand Symbol, 2: 239–250.

Fanelli, A., & Misangyi, V. F. 2006. Bringing out charisma:CEO charisma and external stakeholders.AcademyofManagement Review, 31: 1049–1061.

Fanelli, A., Misangyi, V. F., & Tosi, H. L. 2009. In charismawe trust: The effects of CEO charismatic visions onsecurities analysts. Organization Science, 20: 1011–1033.

Fang, L., & Peress, J. 2009. Media coverage and the cross-section of stock returns. The Journal of Finance, 64:2023–2052.

Finkelstein, S. 1992. Power in top management teams:Dimension, measurement, and validation. Academyof Management Journal, 35: 505–538.

Fiske, S. T., & Taylor, S. E. 2017. Social cognition: Frombrains to culture (3rd ed.). London, U.K.: Sage.

Fiss, P. C., & Zajac, E. Z. 2006. The symbolic managementof strategic change: Sensegiving via framing anddecoupling. Academy of Management Journal, 49:1173–1193.

Foster-Pedley, J., Bond, D., & Brown, R. 2005. The spoken(he)art of strategy—A metaphor analysis of the strat-egydiscourse of 10SouthAfricanbusiness executives.


http://niemanreports.org/articles/staggering-drunks-and-fiscal-cliffs-how-bloomberg-businessweek-uses-metaphors-in-the-news



South African Journal of Business Management,36: 41–54.

Francis, J., & Soffer, L. 1997. The relative informativenessof analysts’ stock recommendations and earningsforecast revisions. Journal of Accounting Research,35: 193–211.

Fuller, R. J., & Metcalf, R. W. 1978. Management disclo-sures: Analysts prefer facts to management’s pre-dictions. Financial Analysts Journal, 34: 55–57.

Fussell, S. R., & Krauss, R. M. 1989. Understanding friendsand strangers: The effects of audience design onmessage comprehension.European Journal of SocialPsychology, 19: 509–525.

Fussell, S. R., & Moss, M. M. 1998. Figurative languagein emotional communication. In S. R. Fussell &R. J. Kreuz (Eds.), Social cognitive approaches to in-terpersonal communication: 113–141. Mahwah, NJ:Lawrence Erlbaum.

Gamson, W. A., & Modigliani, A. 1989. Media discourseand public opinion on nuclear power: A construc-tionist approach.American Journal of Sociology, 95:1–37.

Gao, H., Yu, T., & Cannella, A. A. 2015. The use of publiclanguage in strategy: A multidisciplinary review andresearch agenda. Journal of Management, 42: 21–54.

Gavetti, G., & Rivkin, J. W. 2005. How strategists reallythink—Tapping the power of analogy. HarvardBusiness Review, 83: 54–63.

Gioia, D. A. 1986. Symbols, scripts, and sensemaking:Creating meaning in the organizational experi-ence. In H. P. Sims & D. A. Gioia (Eds.), Thethinking organization: 49–74. San Francisco, CA:Jossey-Bass.

Gioia, D. A., & Chittipeddi, K. 1991. Sensemaking andsensegiving in strategic change initiation. StrategicManagement Journal, 12: 433–448.

Giorgi, S., &Weber, K. 2015.Marks of distinction: Framingand audience appreciation in the context of invest-ment advice. Administrative Science Quarterly, 60:333–367.

Granville, C. 2014. Becoming a financial analyst. Re-trieved from http://www.investopedia.com/articles/financialcareers/06/financialanalyst.asp. AccessedMay23, 2016.

Guo, W. 2014. Executives’ use of emotional language andinvestor reactions. In R. Hart (Ed.), Communicationand language analysis in the corporate world:198–215. Hershey, PA: IGI Global.

Hambrick, D. C., & Finkelstein, S. 1987. Managerialdiscretion: A bridge between polar views of organi-zational outcomes. Research in Organizational Be-havior, 9: 369–406.

Hamington, M. 2009. Business is not a game: The meta-phoric fallacy. Journal of Business Ethics, 86:473–484.

Hatch, M., & Weick, K. 1998. Critical resistance to the jazzmetaphor. Organization Science, 9: 600–604.

Hayward, M., & Fitza, M. 2016. Pseudo-precision? Preciseforecasts and impression management in manage-rial earnings forecasts. Academy of ManagementJournal, 60: 1094–1116.

Healy, P. M., & Palepu, K. G. 1995. The challenges ofinvestor communication. The case of CUC Interna-tional, Inc. JournalofFinancialEconomics,38:111–140.

Healy, P.M., & Palepu,K. G. 2001. Information asymmetry,corporate disclosure, and the capital markets: A re-view of the empirical disclosure literature. Journal ofAccounting and Economics, 31: 405–440.

Heath, C., & Heath, D. 2007. Made to stick: Why someideas survive and others die. NewYork, NY: RandomHouse.

Henry, E. 2008. Are investors influenced by how earningspress releases are written? Journal of Business Com-munication, 45: 363–407.

Heracleous, L., & Jacobs, C. 2008. Understanding organi-zations through embodied metaphors. OrganizationStudies, 29: 45–78.

Ibañez, F., &Hernandez, L. 2011. The contemporary theoryof metaphor: Myths, developments and challenges.Metaphor and Symbol, 26: 161–185.

Janis, I., & Fadner, R. 1965. The coefficient of imbalance. InH. Lasswell, N. Leites, & Associates (Eds.), Languageof politics: 153–169. Cambridge, MA: MIT Press.

Jegadeesh, N., Kim, J., Krische, S. D., & Lee, C. M. C. 2004.Analyzing the analysts: When do recommendationsadd value? The Journal of Finance, 59: 1083–1124.

Johnson, M., & Lakoff, G. 2002. Why cognitive linguisticsrequire embodied realism. Cognitive Linguistics, 13:245–263.

Jorgensen, R. D., & Wingender, J. R. 2004. A survey on thedissemination of earnings information by large firms.Journal of Applied Finance (JAF), 14: 77–84.

Kaplan, S., Murray, F., & Henderson, R. 2003. Disconti-nuities and senior management: Assessing the role ofrecognition in pharmaceutical firm response to bio-technology. Industrial and Corporate Change, 12:203–233.

Keats, B.W., &Hitt, M. A. 1988. A causalmodel of linkagesamong environmental dimensions, macro organiza-tional characteristics and performance. Academy ofManagement Journal, 31: 570–598.

Kindler, J. 2010. To our owners.Pfizer Inc., PfizerAnnualReview 2009: 17–23. New York City, NY: Pfizer Inc.


http://www.investopedia.com/articles/financialcareers/ 06/financialanalyst.asp

http://www.investopedia.com/articles/financialcareers/ 06/financialanalyst.asp

Konig, A., Fehn, A., Puck, J., & Graf-Vlachy, L. 2017. Pri-mary or complex? Towards a theory of metaphoricalstrategy communication in MNCs. Journal of WorldBusiness, 52: 270–285.

Krieger, K., & Ang, J. 2013. The unintended conse-quences of high expectations and pressure on newCEOs. Journal of Business Finance&Accounting, 40:501–526.

Krippendorff, K. 2004. Content analysis: An introductionto its methodology (2nd ed.). Thousand Oaks, CA:Sage.

Kronrod, A., & Danziger, S. 2013. “Wii will rock you!” Theuse and effect of figurative language in consumer re-views of hedonic and utilitarian consumption. Jour-nal of Consumer Research, 40: 726–739.

Lacy, S., & Riffe, D. 1996. Sampling error and selectingintercoder reliability samples for nominal contentcategories. Journalism & Mass Communication Quar-terly, 73: 963–973.

Lakoff, G. 1993. The contemporary theory of metaphor.Metaphor and Thought, 2: 202–251.

Lakoff, G., & Johnson, M. 1980. Metaphors we live by.Chicago, IL: University of Chicago Press.

Lamin,A., &Zaheer, S. 2012.Wall Street vs.MainStreet: Firmstrategies for defending legitimacy and their impact ondifferent stakeholders.OrganizationScience, 23: 47–66.

Landau, M., Meier, B., & Keefer, L. 2010. A metaphor-enriched social cognition. Psychological Bulletin,136: 1045–1067.

Larcker, D. F., & Zakolyukina, A. A. 2012. Detecting de-ceptive discussions in conference calls. Journal ofAccounting Research, 50: 495–540.

Leith, S. 2014. The art of persuasion—How to do folksylikeWarren Buffett. Financial Times. Retrieved fromhttps://www.ft.com/content/68afbbb8-ca14-11e3-ac05-00144feabdc0. Accessed May 23, 2016.

Li, F. 2008. Annual report readability, current earnings,and earnings persistence. Journal of Accounting andEconomics, 45: 221–247.

Lindsley, O. R. 1991. From technical jargon to plain En-glish for application. Journal of Applied BehaviorAnalysis, 24: 449–458.

Liu, D. 2002.Metaphor, culture andworldview. The caseof American English and the Chinese language.Lanham, MD: University Press of America.

Loughran, T., &McDonald, B. 2011.When is a liability nota liability? Textual analysis, dictionaries, and 10Ks.The Journal of Finance, 66: 35–65.

Markarian, F., & Parbonetti, A. 2007. Firm complexity andboard of director compositions. Corporate Gover-nance: An International Review, 15: 1224–1243.

Martens, M. L., Jennings, J. E., & Jennings, P. D. 2007. Do thestories they tell get them themoney theyneed?The roleof entrepreneurial narratives in resource acquisition.Academy of Management Journal, 50: 1107–1132.

Martindale, C. 1990. The clockwork muse: The predict-ability of artistic change. New York, NY: Basic Books.

Matsumoto, D., Pronk,M., & Roelofsen, E. 2011.Whatmakesconference calls useful? The information content ofmanagers’ presentation and analysts’ discussion ses-sions. The Accounting Review, 86: 1381–1414.

McDonnell, M.-H., & King, B. 2013. Keeping up appear-ances: Reputational threat and impression manage-ment after social movement boycotts.AdministrativeScience Quarterly, 58: 387–419.

Medsger, B. 2014. Getting journalism education out of theway. Retrieved from http://journalism.nyu.edu/publishing/archives/debate/forum.1.essay.medsger.html. Accessed May 23, 2016.

Merkl-Davies, D., &Koller, V. 2012. “Metaphoring” peopleout of this world: A critical discourse analysis ofa chairman’s statement of a UK defense firm. Ac-counting Forum, 36: 178–193.

Miles, M. B., & Huberman, A. M. 1994. Qualitative dataanalysis: An expanded sourcebook (2nd ed.). Thou-sand Oaks, CA: Sage.

Miller, A. 2012.The tall ladywith the iceberg: The powerof metaphor to sell, persuade & explain anything toanyone. New York, NY: Chiron Association Inc.

Mintzberg, H. 1973.Anew look at the chief executive’s job.Organizational Dynamics, 1: 20–30.

Mio, J. S. 1997. Metaphor and politics. Metaphor andSymbol, 12: 113–133.

Mio, J. S., Riggio, R., Levin, S., & Reese, R. 2005. Presi-dential leadership and charisma: The effects of meta-phor. The Leadership Quarterly, 16: 287–294.

Mokoaleli-Mokoteli, T., Taffler, R. J., & Agarwal, V. 2009.Behavioural bias and conflicts of interest in analyststock recommendations. Journal of Business Finance& Accounting, 36: 384–418.

Morley, D. 2007. The Cambridge introduction to creativewriting. Cambridge, U.K.: Cambridge University Press.

Oberlechner, T., &Mayer-Schoenberger, V. 2002.Throughtheir own words—Toward a new understanding ofleadership through metaphors. Working Paper No.RWP02–043,KennedySchool ofGovernment, Boston.

Ocasio, W., & Kim, H. 1999. The circulation of corporatecontrol: Selection of functional backgrounds of newCEOs in large U.S. manufacturing firms, 1981–1992.Administrative Science Quarterly, 44: 532–562.

Ortony,A. 1975.Whymetaphors arenecessary andnot justnice. Educational Theory, 25: 45–53.


https://www.ft.com/content/68afbbb8-ca14-11e3-ac05-00144feabdc0

https://www.ft.com/content/68afbbb8-ca14-11e3-ac05-00144feabdc0

http://journalism.nyu.edu/publishing/ archives/debate/forum.1.essay.medsger.html



Partington, A. 1995. A corpus-based investigation into theuse of metaphor in British business journalism, ASp,7–10: 25–39.

Pennebaker, J. W., Booth, R. J., & Francis, M. E. 2007. Lin-guistic inquiry and word count: LIWC 2007. Austin,TX: LIWC.

Petkova,A. P., Rindova,V. P., &Gupta,A.K. 2013.Nonewsis bad news: Sensegiving activities, media attention,and venture capital funding of new technology orga-nizations. Organization Science, 24: 865–888.

Pfarrer, M. D., Pollock, T. G., & Rindova, V. P. 2010. A taleof two assets: The effects of firm reputation and ce-lebrity on earnings surprises and investors’ reactions.Academy of Management Journal, 53: 1131–1152.

Pollock, T. G., & Rindova, V. P. 2003. Media legitimationeffects in the market for initial public offerings.Academy of Management Journal, 46: 631–642.

Pragglejaz Group. 2007. MIP: A method for identifyingmetaphorically used words in discourse. Metaphorand Symbol, 22: 1–39.

Rabe-Hesketh, S., & Skrondal, A. 2008. Multilevel andlongitudinal modeling using Stata. STATA press.

Ramsay, J. 2004. Trope control: The costs and benefits of met-aphor unreliability in the description of empirical phe-nomena. British Journal of Management, 15: 143–155.

Read, S., Cesa, I., Jones, D., & Collins, N. 1990.When is thefederal budget like a baby? Metaphor in politicalrhetoric. Metaphor and Symbol, 5: 125–149.

Rindova, V. P., & Fombrun, C. J. 1999. Constructing com-petitive advantage: The role of firm-constituent in-teractions.StrategicManagement Journal, 20: 691–710.

Roberts, R., &Kreuz, R. 1994.Whydo people use figurativelanguage? Psychological Science, 5: 159–163.

Roush, C. 2011. Show me the money: Writing businessand economics stories formass communication (2nded.). New York, NY: Routledge.

Rozin, P., & Royzman, E. B. 2001. Negativity bias, nega-tivity dominance, and contagion. Personality andSocial Psychology Review, 5: 296–320.

Scheufele, D. A., & Tewksbury, D. 2007. Framing, agendasetting, and priming: The evolution of three mediaeffects models. Journal of Communication, 57: 9–20.

Sedor, L. M. 2002. An explanation for unintentional opti-mism in analysts’ earnings forecasts. The AccountingReview, 77: 731–753.

Shen, W., & Cannella, A. A., Jr. 2002. Revisiting theperformance consequences of CEO succession: Theimpacts of successor type, postsuccession seniorexecutive turnover, and departing CEO tenure. Acad-emy of Management Journal, 45: 717–733.

Shenkar, O., Luo, Y., &Yeheskel, O. 2008. From “distance”to “friction:” Substituting metaphors and redirectingintercultural research. Academy of ManagementReview, 33: 905–923.

Shoemaker, P. J., & Reese, S. D. 1996. Mediating the mes-sage: Theories of influences onmassmedia content.White Plains, NY: Longman Publishers.

Skorczynska, H., & Deignan, A. 2006. Readership andpurpose in the choice of economics metaphors. Met-aphor and Symbol, 21: 87–104.

Sopory, P., & Dillard, J. 2002. The persuasive effects ofmetaphor: A meta-analysis. Human CommunicationResearch, 28: 382–419.

Steen,G. J. 2011.The contemporary theoryofmetaphor—Nownew and improved! Review of Cognitive Linguistics, 9:26–64.

Taylor, S. E. 1981. The interface of cognitive and socialpsychology. In J. H. Harvey (Ed.), Cognition, socialbehavior, and the environment: 189–211. Hillsdale,NJ: Lawrence Erlbaum Associates.

Taylor, S. E. 1991. Asymmetrical effects of positive andnegative events: The mobilization-minimization hy-pothesis. Psychological Bulletin, 110: 67–85.

Thibodeau, P., & Boroditsky, L. 2011. Metaphors we thinkwith: The role ofmetaphor in reasoning.PLoSOne, 6,e16782: 1–11.

Toolan, M. 1988. Narrative: A critical linguistic intro-duction. London: Routledge.

Tsoukas, H. 1991. The missing link: A transformationalview of metaphors in organizational science. Acad-emy of Management Review, 16: 566–585.

Tuchman, G. 1972. Objectivity as strategic ritual: An ex-amination of newsmen’s notions of objectivity. Ameri-can Journal of Sociology, 77: 660–679.

Wade, J. B., Porac, J. F., Pollock, T. G., & Graffin, S. D. 2006.The burden of celebrity: The impact of CEO certifica-tion contests on CEO pay and performance.Academyof Management Journal, 49: 643–660.

Walz, J. 2014. Speak performance: Using the power ofmetaphors to communicate vision, motivate people,and lead your organization to success. New York,NY: Business Expert Press.

Washburn, M., & Bromiley, P. 2014. Managers and ana-lysts: An examination of mutual influence. Academyof Management Journal, 57: 849–868.

Weick, K. 1998. Introductory essay: Improvisation asa mindset for organizational analysis. OrganizationScience, 9: 543–555.

Wenzel, M., & Koch, J. 2017. Strategy as staged perfor-mance: A critical discursive perspective on keynotespeeches as a genre of strategic communication.Strategic Management Journal, 39: 639–663.


Westphal, J. D., & Deephouse, D. L. 2011. Avoiding badpress: Interpersonal influence in relations betweenCEOs and journalists and the consequences for pressreporting about firms and their leadership. Organi-zation Science, 22: 1061–1086.

Westphal, J. D., Park, S. H., McDonald, M. L., & Hay-ward, M. L. 2012. Helping other CEOs avoid badpress social exchange and impression managementsupport among CEOs in communications withjournalists. Administrative Science Quarterly,57: 217–268.

Whittington, R., Yakis‐Douglas, B., & Ahn, K. 2016. Cheaptalk? Strategy presentations as a form of chief execu-tive officer impression management. Strategic Man-agement Journal, 37: 2413–2424.

Wiersema, M. F., & Zhang, Y. 2011. CEO dismissal: Therole of investment analysts. Strategic ManagementJournal, 32: 1161–1182.

Wiesenfeld, B.M.,Wurthmann, K. A., &Hambrick, D. C.2008. The stigmatization and devaluation of elitesassociated with corporate failures: A processmodel. Academy of Management Review, 33:231–251.

Williams, K. 2011. International journalism. London,U.K.: Sage.

Wiseman, 2010.On theuse andmisuse of ratios in strategicmanagement research. In D. D. Bergh & D. J. Ketchen(Eds.), Research methodology in strategy and man-agement, 5: 75–110. Bingley: Emerald Group Pub-lishing Limited.

Wolfe, M. 2012. Putting on the brakes or pressing on thegas? Media attention and the speed of policymaking.Policy Studies Journal: The Journal of the PolicyStudies Organization, 40: 109–126.

Wooldridge, J. M. 2010. Econometric analysis of crosssection and panel data. Cambridge, MA: MIT Press.

Zashin, E., & Chapman, P. C. 1974. The uses of meta-phor and analogy.The Journal of Politics, 36: 290–326.

Zavyalova, A., Pfarrer,M. D., & Reger, R. K. 2016. Celebrityand infamy? The consequences of media narrativesabout organizational identity. Academy of Manage-ment Review 42: 461–480.

Zavyalova, A., Pfarrer, M. D., Reger, R. K., & Shapiro, D. L.2012. Managing the message: The effects of firm ac-tions and industry spillovers on media coverage fol-lowing wrongdoing. Academy of ManagementJournal, 55: 1097–1101.

Andreas Konig ([email protected]) is ChairedProfessor of Strategic Management, Innovation, and

Entrepreneurship at the University of Passau. He receiveda PhD from the University of Erlangen-Nuremberg anda Master of Music in trumpet performance from the RoyalAcademy of Music in London. His research focuses onorganizational transformation, upper echelons theory,and executive communication.

Jan Mammen ([email protected]) is manager for fi-nancial analytics at Beiersdorf AG and lecturer at theUniversity of Erlangen-Nuremberg. He received his PhDfrom the University of Erlangen-Nuremberg. His researchfocuses on corporate restructuring, firm risk, executivecommunication, and artificial intelligence in managementscience.

Johannes Luger ([email protected]) is an assistant professorof strategic management and globalization at the Copen-hagen Business School and an affiliated scholar at theUniversity of St. Gallen. He holds a PhD from the Univer-sity of St. Gallen. Johannes develops and tests modelsbased on behavioral theories to explain organizationalactions and decisions (e.g., investment decisions, resourceallocations, divestitures, etc.).

Angela S. Fehn ([email protected]) currentlyholds a deputy professorship of organization science at theUniversity of Bamberg. She received her PhD from theUniversity of Passau. Her research interests include upperechelons theory, impression management, and inter-disciplinary inquiries at the nexus of the administrativesciences, psychology, and cognitive linguistics.

Albrecht Enders ([email protected]) is a professorof strategy and innovation at IMD, Lausanne. He holdsa PhD in strategic management from the Leipzig GraduateSchool of Management. His major research interest is theresponses of organizations to radical changes in theirenvironment and the role of top executives in theseprocesses.

APPENDIX A

FURTHER ROBUSTNESS CHECKS ANDINFORMATION, AND DATA SOURCES

Additional Robustness Checks

Although our research design addresses numerousconcerns regarding other possible explanations for ourfindings, such as omitted variables, simultaneity, mea-surement errors, and inconsistent inferences (Antonakis,Bendahan, Jacquart, & Lavie, 2010), we ran extensive ad-ditional robustness tests.

In-/exclusion of covariates. Generally, even thoughthe econometric literature indicates that a higher number


mailto:[email protected]





of covariatesmight producemore conservative estimations(Wooldridge, 2010)1, we wished to ensure that our esti-mates were not sensitive to the inclusion of a large numberof covariates.

In particular, apart from the variables removedwhen checking for multicollinearity, we reran Analyses Iand II without firm size, CEO age, CEO tenure, CEO du-ality, contender, outsider, media attention to the CEO,both CEO background variables, CFO involvement, CEOfuture orientation, CEO non-contentual words, and CEOoptimism. In addition, we removed the number of pressreleases and media attention to the firm in Analysis I. InAnalysis II, we removed the debt-to-equity ratio, liquid-ity, cash flow fromoperating activities, abnormal returns,and number of shares traded. The results were unaf-fected in terms of the direction and significance of thecoefficients.

In another check, we reran the models in Analysis Iand included additional variables from Analysis II, suchas the debt-to-equity ratio, liquidity, and operating cashflow. The results were again similar to those of the mainanalysis. Moreover, as the use of ratios or proportionswith similar input variables may cause spurious out-comes (Wiseman, 2010), we dropped those proportionsand still obtained results similar to those of our mainanalysis. Finally, following Benner’s (2010) argumentthat infomediaries might respond negatively if in-cumbents increase investments in discontinuous tech-nologies, we included ameasure of strategic investmentsin the focal quarter (Benner & Ranganathan, 2012): thenatural logarithm of the sum of capital and R&D expen-ditures. In line with conventional practice, we includeda dummy variable, no reported R&D expenditures,which was set equal to 1 for quarters in which R&D ex-penditures were not reported. The results remained thesame.

Individual analysts’ effects. We ran a particularlydetailed check of Analysis II to examine whether indi-vidual effects at the analyst level might provide alter-native explanations for our group-level findings. In thisanalysis, we used multilevel modelling (xtmixed, mlein Stata), and included not only dummies to control forunobservable effects of each analyst and each analyst’s

firm (Deephouse & Heugens, 2009) but also a controlfor time-variant, individual-analyst forecast ability(Fanelli et al., 2009). Given the sizes of our samples, weonly used the EPS-based data on analysts’ favorabilityfor the robustness check at this level.2 In line with thedata structure, we specified a three-level model withmultiple observations over time (level 3) nested withinCEOs (level 2), who were nested within firms (level 1).As crossed effects occur for the time dimension, wefollowed Rabe-Hesketh and Skrondal (2008) in creat-ing an artificial level in which all firms, CEOs, andquarters are nested.3 We added new controls to thoseincluded in our main analysis. First, we controlled forunobservable effects of each analyst and each analyst’sfirm (Deephouse & Heugens, 2009) by includingdummies. These dummies also account for the datastructure. Second, we controlled for each analyst’sforecast ability, assuming that particularly accurateanalysts might be less impressionable (Fanelli et al.,2009). We calculated this variable using the meanforecast error for each analyst in the year of the focalcall:

Mean forecast error of analyst

51N

+N

i51

Actual EPSi 2EPS ForecastiEPS Forecasti

��

where N denotes the total number of one-year EPSforecasts the analyst issued about any firm within thegiven year. Third, we used a dummy variable on theeight-digit GICS level to control for any industry ef-fects. The results of these robustness checks corrobo-rate our findings and can be found in Table A1. The

1 A reader might view testing for sensitivity to the in-clusion of many covariates as particularly important forAnalysis I given the non-significant pairwise correlationbetween the CEOs’ use of metaphorical communicationand the favorability of journalists’ reporting (see Table 2).However, it is crucial to note that a pairwise correlationðrx,y Þ is quasi a priori a biased parameter because it onlyrepresents a scaled regression coefficient

�b1

sxsy

�from the

simple model y 5b0 1b1x1 e. If y is influenced by anyother factor (e.g., z) that is correlated with x, then both rx,yand b1 present biased estimators. Therefore, we disregardthe non-significant pairwise correlation in the test ofHypothesis 1.

2 Given thesignificantly lowernumberof observations forchanges in recommendations (n 5 393 pairs of individualrecommendations before/after the call) than for changes inEPS forecasts (n 5 9,076), the number of predictors in thisrobustness check with multiple dummy variables added istoo high relative to the number of observations of analysts’recommendations to achieve meaningful results. This wasindicated by warnings in Stata’s xtmixed command. How-ever, our hypothesized predictions were supported forrecommendations when we reduced the model’s complex-ity by dropping the dummy variables for analysts, analysts’firms, and industry.

3 As a check, we inverted the structure and used theanalysts, their firms, and time as levels while control-ling for CEOs and firms with dummies. The results werenot affected by this change in model specification. No-tably, a higher-order model with five levels (i.e., CEO,firms, analysts, employers, and time) would be prob-lematic, as there would be additional crossed effectsbetween analysts and CEOs (e.g., analystswho issue EPSforecasts for different firms in our sample in the samequarter).


TABLE A1Results of Multilevel Analysis of CEOs’ Use of Metaphorical Communication on the Favorability of

Individual Analysts’ Evaluationsa


Prior firm performance change 0.39*** 0.42*** 0.42***(0.11) (0.11) (0.11)

Prior firm performance volatility 0.07 0.07 0.08(0.14) (0.14) (0.14)

Negative earnings surprise† –1.45*** –1.46*** –1.78***(0.09) (0.09) (0.10)

Positive earnings surprise† 0.55*** 0.54*** 0.54***(0.04) (0.04) (0.04)

Firm size 0.03 0.04 0.04(0.04) (0.04) (0.04)

Number of press releases 0.00* 0.001 0.001

(0.00) (0.00) (0.00)Media attention to the firm –0.00 –0.00 –0.00

(0.00) (0.00) (0.00)Debt–to–equity ratio 0.011 0.011 0.02*

(0.01) (0.01) (0.01)Dividends per share –0.02 0.01 –0.07

(0.15) (0.15) (0.15)Liquidity –0.02 0.02 0.07

(0.13) (0.13) (0.13)Cash flow from operating activities –0.011 –0.011 –0.011

(0.01) (0.01) (0.01)Abnormal returns 0.02*** 0.02*** 0.02***

(0.00) (0.00) (0.00)Number of shares traded 0.00*** 0.00*** 0.00*

(0.00) (0.00) (0.00)CEO age –0.00 –0.00 –0.00

(0.01) (0.01) (0.01)CEO tenure 0.00 0.00 0.00

(0.01) (0.01) (0.01)CEO duality –0.101 –0.08 –0.08

(0.06) (0.06) (0.06)Contender –0.08 –0.09 –0.10

(0.12) (0.12) (0.12)Outsider 0.09 0.10 0.09

(0.12) (0.12) (0.12)Media attention to the CEO –0.00 –0.00 0.00

(0.00) (0.00) (0.00)CEO celebrity –1.21*** –1.23*** –1.19***

(0.27) (0.27) (0.27)CEO background (Sales & Marketing) –0.09 –0.08 –0.08

(0.09) (0.09) (0.09)CEO background (Finance) 0.60*** 0.60*** 0.60***

(0.08) (0.08) (0.08)Number of analysts following –0.01* –0.01* –0.01**

(0.00) (0.00) (0.00)Mean forecast error –0.03 –0.03 –0.03

(0.02) (0.02) (0.02)Share of CEO’s words 0.06 0.05 0.09

(0.15) (0.15) (0.15)CFO involvement –0.021 –0.03* –0.021

(0.02) (0.02) (0.02)CEO future orientation –0.371 –0.39* –0.47*

(0.19) (0.19) (0.19)CEO image-based language –1.37 –1.01 –0.55

(1.17) (1.17) (1.16)


sequence of the models follows the same logic as ourmain analyses.

Additional controls. In another robustness checkfor Analysis II, we accounted for the idea that analysts’favorability could be affected by whether managementprovides forecast guidance during the conference call.In this regard, we followed recent research (e.g., Chen,Crossland, & Luo, 2015) in collecting data from FirstCall’s Company Issued Guidelines database (CIG). Wecreated a binary variable that took a value of “1” if a firmin our sample issued forecast guidance on the date ofa conference call. We included this variable in AnalysisII as an additional control, with results similar to thosereported in our main analysis.

Even though CEOs rarely use stories as a rhetoricaltool in conference calls (and, if so, almost exclusively inthe Q&A part of the call), we also re-ran our models witha control for the number and length of stories told byCEOs. In contrast to metaphors, stories a priori featureagonists that are part of a sequence of interrelated events(Toolan, 1988) and they have been found to be influen-tial in a financial-market context (Martens, Jennings, &Jennings, 2007). However, we observed no change in theresults.

In other robustness checks, we included dummy vari-ables for CEOs’ prior experience in production and oper-ations, technologyR&Dand science, legal and compliance,human resources, and strategy. In addition, we tested forCEOs’ educational backgrounds by including dummyvariables for CEOs’ degree (i.e., MBA, BS/MSc, BA/MA,

LLB/JD, CPA/CFA/CMA, and PhD/MD). Our main resultswere unaffected.

Simultaneity. Simultaneity concerns might be rela-tively non-critical in our study because CEOs are rela-tively unlikely to be able to anticipate changes in theevaluations of journalists or analysts after a given con-ference call. Nevertheless, inAnalyses I and II, we testedfor the possibility that CEOs might use more metaphor-ical language during a conference call in response tofavorable journalist reporting before a conference call.We did so by regressing CEOs’ use of metaphoricalcommunication on the respective measures of favor-ability of infomediaries’ reportingt-1. After controllingfor the same sets of variables included in AnalysesI and II, respectively, we did not find any indicationsthat prior favorability in infomediaries’ reporting signifi-cantly determines CEOs’ use of metaphorical communica-tion.

Endogeneity.One particularly important alternativeexplanation for our results could be that CEOs’ use ofmetaphorical communication might be driven, in part,by the same factors that also affect infomediaries’favorability. Such endogeneity is important because itinherently limits our ability to make causal claims basedon our data. Moreover, given the current state ofknowledge on metaphorical communication, it is im-possible to entirely rule out such endogeneity. Apartfrom marginally related systematic evidence, no re-search exists that allows us to develop a clear and com-prehensive theoretical model of the drivers of CEO’s use

TABLE A1(Continued)


CEO comprehensibility 0.06*** 0.06*** 0.06***(0.01) (0.01) (0.01)

CEO numerical language 4.65* 4.65* 5.97**(2.18) (2.18) (2.17)

CEO non-contentual words 0.12 0.07 0.11(0.15) (0.15) (0.15)

CEO optimism 0.33*** 0.37*** 0.30***(0.07) (0.07) (0.07)

CEO’s use of metaphorical communication†–7.24*** –6.85***(1.71) (1.72)

CEO’s use of metaphorical communication –75.47***x Negative earnings surprise (14.19)

CEO’s use of metaphorical communication 34.92***x Positive earnings surprise (4.73)

N 6,969 6,969 6,969Log likelihood –7913.52 –7904.63 –7852.94LR x2 against null model 1819.30*** 1842.13*** 1973.74***

1p , .1; * p , .05; ** p, .01; *** p , .001.aBased on EPS–forecast measure of analysts’ favorability. Standard errors in parentheses. Models include dummies for each quarter from

2002 to 2011, 8–digit GICS levels, and analyst firms.†Variable is centered at its mean.


of metaphorical communication.4 As such, developinga baseline first-stage model of the drivers of the use ofmetaphorical communication is outside the scope of ourpaper. Furthermore, we are unable to randomly assign“metaphorical communication” across CEOs through anexperiment. However, to at least tentatively scrutinizewhether such endogeneity taints our results, we exam-ined the outcomes of our simultaneity analyses andfound that no theoretically intuitive factors includedin our data set, such as strategic change (Benner& Ranganathan, 2012; Kotter, 1996), significantly pre-dicted CEOs’ use of metaphorical communication.Moreover, considering that CEOs might feel tempted touse metaphorical communication in complex circum-stances, we regressed CEOs’ use of metaphorical com-munication on twomeasures of complexity thatwere notincluded in our set of controls: (1) external complexity inthe form of munificence, instability, and complexity5

(Keats & Hitt, 1988); and (2) internal complexity asmeasured by the number of business segments in the

focal firm (Markarian & Parbonetti, 2007). None of thesevariables had a significant influence on CEOs’ use ofmetaphorical communication (at p , 0.1), regardless ofwhether we included our existing control variables inthe models.

In an additional attempt to cope with potential endoge-neity, we followed Wiersema and Zhang (2011). Morespecifically, for Analyses I and II, we first regressed CEOs’use of metaphorical communication on all control vari-ables in the respectivemodels and then calculated residualvalues of CEOs’ use of metaphorical communication.6 Wethen reran Analyses I and II, replacing the observed valuesof CEOs’metaphorical communication with the residuals.

TABLE A2Data Sources and Corresponding Data / Variables

Thomson Reuters IBES (Institutional Brokers Estimates System)- Favorability of analysts’ evaluations- Negative earnings surprise- Positive earnings surprise- Number of analysts following

Factiva (Dow Jones)- Favorability of journalists’ reporting- Media attention to the firm- Media attention to the CEO

Thomson Research and Seeking Alpha- Transcripts of quarterly earnings conference calls

Compustat- Prior firm performance change- Prior firm performance volatility- Firm size- Debt-to-equity ratio- Dividends per share- Liquidity- Cash flow from operating activities

Marquis Who’s Who; Publicly available information (annualreports, company information, Bloomberg Executive Profilesand Biography)

- CEO age- CEO tenure- CEO duality- Contender- Outsider- CEO background (Sales & Marketing)- CEO background (Finance)

www.stevieawards.com- CEO celebrity

Business Wire- Number of press releases

Eventus (based on CRSP data)- Abnormal returns

Center for Research and Security Prices (CRSP)- Number of shares traded

4 We conducted an extensive, systematic review of theliterature in cognitive linguistics and related domains todetermine whether it was possible to derive a theoreticalmodel predicting CEOs’ use of metaphorical communica-tion. Overall, we concluded that a full, generalizable pic-ture of the drivers of metaphorical communication islacking from a linguist’s perspective. Most of the work weidentified indicates that certain contextual factors couldinfluence the frequency of metaphor use, especially: (1)whether one communicates with oneself (“inner speech”)rather than with others (Fussell & Krauss, 1989); (2)whether the communicator talks about emotions, espe-cially sadness, rather than behavior (e.g., Fainsilber &Ortony, 1987; for overviews, see Fussell &Moss, 1998, andKronrod & Danziger, 2013); (3) whether the communicatortries to integrate unfamiliar perspectives (Corts & Pollio,1999); (4) the purpose of the communication (e.g., whetherthe speaker aims to make a speech more interesting or toclarify an issue (Roberts & Kreuz, 1994); and (5) whetherthe communicator is addressing a general or specialistaudience (Skorczynska & Deignan, 2006). We do not seeany obvious, non-speculative reason why these constructsshould bias our results. Notably, the fact that we arestudying a relatively homogeneous group of CEOs in a rel-atively homogeneous setting (i.e., conference calls) shouldaccount for some of these factors (e.g., the purpose of thecommunication or the intended audience). Moreover,many of the theories and results found in prior research arehardly generalizable to the CEO-infomediary context (e.g.,Fussell & Kreuz, 1998). While we understand that thisfinding does not rule out endogeneity, it does make usmore confident that we did not miss fundamentally im-portant antecedents of metaphorical communication thatwould bias our results.

5 We tested all elements separately and as a factor scoreof all three elements.

6 To avoid an arbitrary choice of predictors in the first-stage regression, we also ran the endogeneity test withdifferent subsets of predictors. All of them had robustresults.


http://www.stevieawards.com

As such, we tested whether the component of CEOs’ met-aphorical communication that was uncorrelated with ourcontrol variables had a significant effect on infomediaries’favorability. In support of our findings, the coefficient ofthe residuals was positive and significant (p , 0.05) forAnalysis I, and negative and significant (p , 0.05) forAnalysis II.

Other checks. We also tested calendar-year dummiesand quarter dummies instead of a dummyvariable for eachunique quarter in our dataset. Our results remained robust.Moreover, we tested whether the moderating effects ofnegative earnings surprises were due to outliers by win-sorizing the negative earnings surprises variable at the 1percent and 5 percent levels. Our results were not affectedby those changes. Further, while the hypothesis that allfirm fixed effects are zero was strongly rejected, we recal-culated our models using pooled regression with standarderrors clustered at the firm level and found consistentresults.

Finally, we used change scores for our dependent vari-ables because these scores are “regardedapowerful tool formaking causal inferences with nonexperimental data”(Allison, 1990: 93). Despite these advantages, changescores have been criticized for producing inaccurate re-sults, mostly due to potential reliability concerns andconcerns related to regression toward the mean (Bergh &

Fairbank, 2002). In order to account for both concerns, weimplemented regressor variable models (Allison, 1990) inAnalyses I and II, which yielded results similar to thoseobtained from the change-score models.

Additional Information on CEOs’ Use ofMetaphorical Communication

Readers might be interested in more descriptive detailson CEOs’ use of metaphorical communication. In thisregard, fourteen CEOs, used it in two percent or more oftheir contentual communication. Fourteen CEOs in oursample did not use metaphorical communication at all.The remaining CEOs are approximately normally distrib-uted between these two values. Not surprisingly, the morecontentual words a CEO speaks in a conference call, thehigher the overall number of metaphorical words in thecall (correlation of 0.56). Finally, there does not seem toexist a time trend regarding CEOs’ use of metaphoricalcommunication, as it does not systematically increase ordecrease along a CEO’s tenure.

Overview of Data Sources

Table A2 summarizes the data sources for all variables.


Copyright of Academy of Management Journal is the property of Academy of Managementand its content may not be copied or emailed to multiple sites or posted to a listserv withoutthe copyright holder's express written permission. However, users may print, download, oremail articles for individual use.

Documents

Silver Bullet or Ricochet? CEOs’ Use of Metaphorical ......Silver Bullet or Ricochet? CEOs’ Use of Metaphorical Communication and Infomediaries’ Evaluations Andreas König, Jan